Market sentiment of electric vehicles in Malaysia

Abstract

Electric Vehicles (EVs) are gaining traction globally due to their potential for energy efficiency and emission reduction. In Malaysia, where the government is actively promoting EV adoption, market sentiment plays a critical role. Most existing studies on EV market sentiment in Malaysia rely heavily on survey-based methods, which may be limited in both scale and objectivity. This study applies text mining techniques, specifically topic modelling and sentiment analysis to examine Malaysian consumers’ attitudes toward EVs using user comments from a major national automotive information website. A total of 9,618 user comments between January 2023 and February 2024 were analysed using LDA/VADER. The results indicate that “price” is the most frequently mentioned keyword, while key discussion topics include market trends, government incentives, battery and charging, and product features. Sentiment analysis suggests that consumers hold a mildly positive or curious attitude toward EVs. These findings provide practical insights for policymakers seeking to enhance market engagement and accelerate EV adoption in Malaysia.

Introduction

Addressing climate change and reducing carbon emissions through sustainable transportation is a crucial global priority, aligning with the Sustainable Development Goal (SDG) 13—Climate Action. The adoption of Electric Vehicles (EVs) represents a significant strategy to mitigate environmental impacts while advancing technological innovations in the automotive industry. EVs encompass Pure Electric Vehicles (PEVs) and Hybrid Electric Vehicles (HEVs) where the former runs solely on electric power and the latter operates on a combination of electric and gasoline. These vehicles offer a sustainable alternative to traditional internal combustion engine (ICE) vehicles¹. Compared to conventional vehicles, EVs potentially enhance fuel consumption efficiency by 40 to 60% and reduce carbon dioxide emissions by 30 to 50%^2,3.

The nation’s commitment to achieving its target of net-zero greenhouse gas (GHG) emissions by 2050 is ambitiously concretise in the introduction of Malaysia’s National Energy Transition Roadmap (NETR). Albeit this vision, the current market penetration of EVs in Malaysia remains relatively low. According to the Malaysian Automotive Association (MAA)⁴, EVs charted a national sales volume of 38,055 units in 2023, accounting for only 4.76% of the Total Industry Volume (TIV). This figure falls short of the government’s targets of 20% by 2030 and 80% by 2050. Bridging this gap is essential for realizing Malaysia’s environmental and economic objectives, particularly in the context of climate action and sustainable mobility.

Government support plays a crucial role in accelerating EV adoption and positioning Malaysia as a regional hub for the EV production and consumption⁵. Efforts to bolster the use of EVs have kickstarted with initiatives, such as financial incentives for purchasers and the establishment of charging infrastructure. However, albeit these measures, a significant challenge remains—consumers’ acceptance and their willingness to embrace the transition from conventional vehicles.

Past research identified socio-psychological factors including awareness, interest and perception as key factors shaping the adoption of EVs as an emerging technology^6–8. While these studies provide valuable insights, most rely on structured surveys that constrain responses to predefined constructs, making it difficult to capture the broader socio-structural concerns, contextual anxieties, or equity-related perceptions that may also influence EV uptake. A thorough understanding of consumer sentiments and attitudes is therefore crucial not only for addressing behavioural barriers but also for identifying underexplored dimensions of public discourse that can inform more socially responsive EV policies⁹.

The rise of social media and technological innovation has enabled the generation of extensive user-generated data. Combined with advanced Natural Language Processing (NLP) methods, these platforms allow researchers to observe large-scale, organic conversations in real time, offering a rich and less constrained data source compared to traditional surveys¹⁰. This capability enables the detection of evolving narratives, emotional tendencies, and context-specific concerns that may not emerge through conventional data collection methods¹¹. Through text mining and sentiment analysis, decision-makers can better understand users’ feelings, needs and expectations by analysing unstructured content such as customer reviews and social media posts, thereby revealing latent issues and emerging public expectations that are critical for informed EV planning and policy design¹².

Moreover, previous studies on EV adoption in Malaysia remain limited, often focusing on questionnaires as a methodology. This is evident in Sang and Bekhet’s¹³ study where they proposed a model to explore Malaysians’ usage intention for EVs. Their findings reported that factors such as social influence and government intervention are related to EV adoption in the country. Subsequently, similar studies were also conducted by Adnan et al.¹⁴, Tanwir and Hamzah¹⁵, and Asadi et al.¹⁶. Questionnaire and survey methods applied in these studies involve volunteer and self-selection bias which may pose a challenge since self-selecting participants may not accurately represent the broader target population¹⁵. To overcome this limitation, recognizing the role of social media as a source of information offers another avenue to assess public awareness levels¹⁷. A recent study by Hamzah et al.¹⁸ utilised content analysis to investigate consumer perceptions of Malaysian HEV adoption. However, this method requires careful reading and manual identification of topics, lacking objectivity. To overcome these weaknesses, this study employs Python to perform text mining including word frequency statistics, topic modelling, and sentiment analysis to analyze comments on electric and hybrid vehicles articles from the Paultan.org website (This Malaysian website is a credible automotive news platform with high user engagement¹⁸) Sentiment analysis and topic modelling conducted on social media have been widely used in EV market research to identify and prioritized the topics of greatest concern to the public¹⁹. However, there is a lack of comparable research in Malaysia.

The significance of this study lies primarily in its contextual contribution and its empirical identification of Malaysia-specific drivers of EV sentiment. While text mining techniques such as LDA and VADER have been widely used in prior sentiment studies, their application to large-scale, unsolicited social media discussions on EV adoption in Malaysia remains limited. This study therefore does not claim methodological novelty in the development of new algorithms; rather, it demonstrates the value of applying established text mining tools to an under-researched emerging-market context where consumer perceptions are shaped by distinctive economic, infrastructural, and policy conditions.

In this study, public online discussion is treated as a window into how consumers evaluate EVs as a relatively new mobility innovation. This perspective is informed by Rogers’²⁰. Diffusion of Innovation theory (DOI), which suggests that adoption depends on how potential users perceive an innovation’s advantages, fit with existing routines and institutions, ease or difficulty of use, opportunity for practical experience, and visibility in the wider social environment. These dimensions are particularly relevant to EV adoption because consumers do not evaluate EVs only as vehicles, but also as part of a broader system involving prices, brands, charging infrastructure, government incentives, road-tax rules, and everyday mobility habits.

In doing so, the study contributes to the EV adoption literature in two ways. First, it extends existing research beyond the developed-market and China-focused settings that dominate prior studies, providing evidence from Malaysia, where EV adoption is still at an early stage and where public discourse reflects concerns that may not be captured by structured surveys. Second, the findings show that Malaysian EV sentiment is shaped not only by conventional factors such as cost, charging infrastructure, and environmental awareness, but also by policy ambiguity and concerns over distributive fairness. These issues are particularly salient in an emerging-market context, where questions about incentives, affordability, and equitable access can strongly influence public acceptance of EV transition policies. By highlighting these context-specific drivers and barriers, this study offers a more grounded understanding of EV adoption in Malaysia and provides insights that may be relevant to other emerging economies facing similar policy and infrastructure challenges.

This paper is structured as follows: Section “Literature review” reviews the relevant literature. Section “Research methodology” outlines the methods of data collection and analysis, while Section “Results and discussion” presents the analysis and discussion of the results. Section “Conclusion” provides the conclusions and policy implications. Finally, Section “Limitations and directions for future research” discusses the limitations of the study and offers directions for future research.

Literature review

Market sentiments toward EVs

Studies concerning consumer’s intentions of EV adoption (CAVIE) focus on four main aspects: demographical, situational, contextual, and psychological. These studies pay attention to factors such as environmental concerns, technology, financial considerations, market dynamics, policy measures, charging infrastructure, as well as perception and attitudes towards EVs²¹. Krishna²² argues that many survey-based studies emphasise EV’s physical factors such as range and safety but overlook the importance of abstract elements such as sentiment and desirability. Several qualitative research has identified several challenges related to the adoption of EVs in Malaysia including charging infrastructure, market pricing, environmental awareness, pandemic and duty regularity^5,23,24. It is important to note that market sentiments have not been adequately considered in these studies. Furthermore, related issues and topics that garner market attention in Malaysia have also been neglected.

To ensure higher market acceptance of EVs, it is necessary to address technical issues i.e., battery capacity and vehicle weight as well as consumers’ needs and concerns²⁵. When it comes to consumers, human sentiment is found to influence the decision-making process at both an individual and group level²⁶. In this sense, understanding market sentiment and responses when it comes to technological adoption and acceptance can inform policymakers and technology developers, simultaneously enhancing their relationship with the public²⁷. For consumers, findings from sentiment analysis are essential to aid them in making informed decisions, thus, promoting the expansion of emerging EV markets including the Malaysian market²⁸.

Research from around the world has found that user sentiment, feelings and emotion as significant factors influencing the adoption of EVs^29–33. Positive user sentiment and attitude are important determinants for the widespread adoption of EVs, while negative sentiment inevitably discourages EV purchases^34,35. Malaysian research on the uptake of EVs highlights the extent of consumer perception, feelings, and subjective social norms in shaping behavioural intentions toward EVs. These studies highlight the psychological and social elements underlying these choices^13–16.

A limitation observed within these studies is that it is often limited to a single Malaysian city or region such as the works of Mohiuddin et al.³⁶ as well as Noor and Wen³⁷ who explored the purchasing intention of EVs among Kelantan university students and people in the Klang Valley area respectively. Most of these studies rely predominantly on surveys and questionnaires and are inherently subject to methodological limitations such as limited sample sizes and predefined questions.

Text mining approach using online platform data

An online automotive platform is a digital service that enables activities related to vehicles, such as buying, selling, accessing services, and sharing information, over the Internet. It connects consumers, dealers, and car enthusiasts within the automotive industry. Within the automotive industry, user opinions and feedback on social media or online forums are generally perceived as more credible and trustworthy than those made by companies²². Textual information on such platforms can be analysed using text mining for various purposes, including product planning and marketing³⁸. This information, commonly known as User-Generated Content (UGC)³⁹, is often collected and analysed by a significant majority of companies through various methods on social media to retain brand health⁴⁰. According to the DOI Theory, social networks can improve the use of innovations in early adoption phases such as EVs⁴¹. These networks encourage rapid information dissemination, enable direct observation of feedback, and amplify strong opinions, thereby shaping positive public perception.

An online automotive platform refers to a digital environment that facilitates vehicle-related activities, including information exchange, purchasing decisions, service access, and peer-to-peer discussion among consumers, dealers, and automotive enthusiasts. Within the automotive domain, opinions expressed on online platforms and social media are often perceived as more credible than firm-generated communications, as they reflect experiential knowledge and peer evaluation rather than promotional intent²². Consequently, user-generated content (UGC) has become an increasingly important data source for understanding consumer perceptions, market sentiment, and behavioural responses^38,39.

The flexible nature of text mining holds great potential as a retrieval and analytical tool for large-scale and complex data⁴². It facilitates the extraction of critical insights from various text data including search engines, customer relationship management systems, product review analysis and social media analytics. This feature enables sophisticated opinion mining, feature extraction, sentiment, predictive and trend analysis⁴³. Integrated with web crawlers, the range of accessible data for this method can be expanded besides increased objectivity¹⁹. What matters is that, beyond its technical flexibility, text mining captures forms of public expression that are systematically under-represented in structured questionnaires. Prior computational social science research suggests that online discourse often includes sarcasm, distrust of institutions, anecdotal risk-taking, and persistent misconceptions, which are difficult to capture using predefined survey constructs or Likert-scale measures^44,45.

Text mining is utilised in sentiment analysis for market forecasting, owing to its interdisciplinary background that includes linguistics, machine learning, and behavioural economics⁴⁵. Recent studies have employed text mining to examine electric vehicle (EV) adoption in countries such as China^19,28,46, India²⁵, and the United States^47,48. These scholarly works offer valuable insights into public sentiment, topic evolution, and policy discourse, and have provided actionable recommendations for accelerating EV deployment in each context.

As an emerging market, ASEAN has received limited attention in the field of text mining applied to EV adoption. Nugroho and Widianto⁴⁹ analyze Twitter data in Indonesia using Aspect-Based Sentiment Analysis and zero-shot classification techniques. Twitter is well suited to capturing immediate public reactions to policy announcements and emerging issues. However, such short-form and event-driven discourse may place greater emphasis on momentary reactions than on sustained, decision-oriented discussions related to EV adoption. Similarly, Emphan et al.⁵⁰ analyze YouTube comments in Thailand using supervised machine learning models, including SVM, to classify public sentiment toward EVs. Although the study demonstrates technical rigor in model optimization, its contribution to understanding EV adoption is limited by a weak linkage between sentiment classification and adoption-relevant mechanisms. The analysis does not distinguish between affective reactions, policy critiques, and practical adoption barriers, thereby treating sentiment as an outcome rather than as an explanatory factor within a broader behavioral or socio-technical framework. In contrast, the present study draws on user-generated comments from a specialized automotive platform widely used in Malaysia. Discussions on this platform are typically longer, more technical, and more closely aligned with purchasing and policy considerations, including pricing, taxation, and charging compatibility. This enables a more direct examination of how concerns and sentiments relate to persistent barriers to EV adoption in the Malaysian context.

A small number of recent Malaysian studies have begun to apply text classification and sentiment analysis to online platforms—for example, analysing YouTube comments to identify dominant themes like battery life, charging station availability, and price⁵¹. Yet these studies typically short of deeper techniques such as topic modelling, temporal analysis, or event-based sentiment tracking. By contrast, international research has shown that media attention can significantly influence EV adoption trajectories⁵², and that public discussion of issues like range anxiety, infrastructure gaps, and policy confusion can be tracked over time to reveal patterns not accessible through surveys^53–55.

Survey-based studies on EV adoption typically focus on demographic factors, attitudes, and concerns, but they are often limited by small sample sizes, regional biases, and the inability to track changes in sentiment over time^36,37. While surveys provide valuable structured insights, they tend to overlook emerging issues and evolving public perceptions. In contrast, text mining allows for continuous, large-scale analysis of user-generated content from platforms like social media. This approach offers real-time insights into public sentiment and reveals dynamic shifts in consumer concerns, including charging infrastructure, government policies, environmental issues, as well as spontaneous worries and common misunderstandings. These aspects are typically difficult to capture through conventional survey instruments with fixed response structures.

To sum up, text mining offers advanced techniques such as topic modeling and temporal sentiment analysis, which can uncover emerging trends and track the evolution of public opinions over time^52,54,55. These techniques provide a deeper understanding of consumer attitudes that surveys are unable to capture. In the context of ASEAN, where market dynamics are rapidly evolving, particularly in countries like Malaysia, text mining becomes a powerful tool to track the shifting sentiments towards EV adoption.

Research methodology

We utilised a web crawler in our study to obtain user comments on news posts about EVs on an authoritative Malaysian automotive online platform before conducting a textual analysis. The analytical process includes focusing on market concerns, and topics related to consumers’ attention as well as sentiment analysis. Our research design is conceptualised in Fig. 1.

Fig. 1.

Research design.

Data collection

We conducted an exploratory analysis of publicly available data with a specific focus on UGC found on Paultan.org. As mentioned earlier, this website is a leading and credible automotive Malaysian news website with a high user engagement level¹⁸. As of April 2024, their Instagram account boasted over 158,000 followers, and their Facebook public page has amassed more than 1 million followers. Therefore, this platform can be seen as a reliable resource to obtain textual data on the Malaysian population’s attention and sentiment towards EVs.

Although the website features a dedicated section for EVs, further investigation indicates that not all EV-related news is included in the dedicated section. Therefore, the site’s search function was utilised to gather information concerning EVs, such as EV launch events, EV policies, EV reviews, and EV industry news. Additionally, the selected information also accounts for different types of EVs including Pure Electric Vehicles (PEVs), Battery Electric Vehicles (BEVs), Hybrid Electric Vehicles (HEVs), and Plug-in Hybrid Electric Vehicles (PHEVs).

The unit of analysis for this study is a single user comment, following the approach outlined by Holtz et al.⁵⁶. Data collection was conducted using a web crawler developed in Python 3.12.0, enabling large-scale extraction of user-generated content⁵⁷. Initially, 13,879 comments were mined from 1077 EV-related articles published between January 1, 2023, and February 29, 2024. After applying manual filtering based on predefined criteria, the final dataset consisted of 9,618 comments extracted from 892 articles. The manual filtering was conducted based on the following criteria:

1. Articles not related to EVs irrelevant articles were removed to ensure that the analysis remains focused on the target subject matter;

2. Comments not related to EVs off-topic comments, even if attached to EV-related articles, were excluded to preserve thematic consistency;

3. Blank comments, garbled characters, entries consisting solely of interjections (e.g., “Wow”, “Haha”), or web links—such content lacks semantic value and would not contribute meaningfully to sentiment analysis;

4. Comments not primarily written in English only English-language comments were retained, as the scope of this study is limited to English discourse on EVs. This boundary was set to maintain consistency in linguistic context and allow more focused interpretation of sentiment patterns within a specific language environment.

Text mining

Text mining is the process of extracting meaningful information from unstructured or semi-structured textual data which aids decision-making and strategic planning by revealing patterns, trends, and insights. This research focuses primarily on topic modelling and sentiment analysis to explore and interpret large datasets. We utilised topic modelling to uncover underlying themes within texts and sentiment analysis to explore further expressed emotions and opinions. The combination of these two approaches has been proven effective across various research domains such as involving government service⁵⁸, healthcare⁵⁹, tourism⁶⁰, folk beliefs⁶¹, financial reputation risk⁶² and global warming issues⁶³. The following sections discuss the application of these two methods:

Data preprocessing

In our study, we employed a streamlined data preprocessing approach, enhancing the relevance and clarity of text data for analysis. This process involved:

1. Natural Language Toolkit (NLTK)⁶⁴ for Text Processing: The NLTK library was utilised to perform lemmatisation which converts all text to lowercase, removes punctuation and pure numbers, filters out stop words (filler words and other daily vocabulary unrelated to the core content of the text), and tokenise the text. Frequently adjacent words are also converted into compound terms using hyphens to prevent misinterpretation of phrases.

2. Regular Expressions for Data Transformation: Regular expressions were identified to eliminate web URLs and transform numerical expressions into meaningful terms. For example, converting “RM” and a number to “Price”(e.g., “RM100K” = “Price”), and numerical values followed by “kw/h” to “Battery Capacity”⁵,. This process transforms the data to be more insightful and align better without aim.

3. Established a thesaurus to restore abbreviations and local slang words, replacing terms like “govt” with “government”, and combining near-synonyms into more formal terms (e.g., “cheap” to “affordable”). Additionally, the Langid library⁶⁵ was also utilised to filter out non-English terms, ensuring the dataset contained exclusively English words.

The preprocessing rules and a sample cleaned comment are provided in Appendix 1.

LDA topic modelling

This process involves extracting abstract “topics” from a document that can be used for opinion mining and text classification⁶⁶. There are several topic modelling techniques but the most common technique is the Latent Dirichlet Allocation (LDA)⁵⁹. The LDA’s most notable benefit is its ability to extract topics from document collections without the need for prior knowledge input⁶⁷. It conceptualises each document as a mixture of various topics, characterising each topic by a distribution over words. It operates by assuming that the generation of each document involves a topic distribution selected from a Dirichlet distribution where words are then drawn from these topics’ word distributions. Blei et al.⁶⁸ and Maier et al.⁶⁹ expounded on the LDA model comprehensively in their works respectively. The LDAvis package⁷⁰ is also applied in Python to visualise the results of LDA topic modelling for a more intuitive discussion.

Sentiment analysis

The Valence Aware Dictionary and Sentiment Reasoner (VADER) sentiment analysis tool⁷¹ is employed in this study. VADER is a lexicon and rules-based sentiment analysis tool tailored specifically for sentiments expressed on social media. This tool recognises commonly used terms, phrases, abbreviations and jargon simultaneously also accounting for grammatical structures commonly used in the vernacular of social media platforms such as punctuation, negation, hedging and magnification⁷². It draws on a list of sentiment lexical features (e.g., words such as ‘love’, ‘hate’, ‘bad’) that are commonly identified according to their semantic orientation as either positive or negative. VADER scores these positive and/or negative orientations and measures the intensity of these sentiments. i.e., differentiating between varying degrees of positivity between words like ‘good’, ‘better’, and ‘best’.

Combining both the LDA model and VADER sentiment analysis, allows us to extract themes and sentiment orientations from large volumes of text data. Consequently, these findings provide a better understanding of EV adoption in Malaysia through accurate evidence for policy formulation and market strategies.

Results and discussion

General market concern

In the first phase of analysis, we focused on statistical word frequencies (see Table 1), which present the top 60 most frequently mentioned words in the dataset. To complement this, we also generated a word cloud (Fig. 2)—a visual tool that displays words in varying font sizes based on their frequency. The larger the font size, the more frequently the word appeared in the data. This visualisation offers an intuitive snapshot of the most salient terms and key topics discussed in the text.

Table 1.

Word frequency statistics.

Rank	Word	Frequency	Rank	Word	Frequency	Rank	Word	Frequency
1	price	2628	21	drive	379	41	bad	228
2	vehicle	2278	22	toyota	378	42	fuel	227
3	electric-vehicle	2277	23	model	359	43	sale	222
4	malaysia	1199	24	cost	350	44	road	220
5	government	904	25	country	342	45	charger	202
6	china	789	26	bmw	342	46	fire	202
7	people	684	27	proton	339	47	geely	200
8	battery	667	28	tax	336	48	afford	200
9	buy	657	29	market	318	49	change	185
10	tesla	628	30	design	316	50	quality	183
11	charge	591	31	money	310	51	japan	183
12	affordable	525	32	comparison	295	52	ugly	179
13	charge-time	494	33	petrol	288	53	nice	179
14	sell	479	34	pay	283	54	base	176
15	brand	438	35	power	282	55	poor	176
16	expensive	429	36	road-tax	274	56	manufacturer	176
17	mercedes	429	37	hybrid	266	57	thailand	174
18	range	412	38	engine	262	58	tech	172
19	ice-vehicle	410	39	company	241	59	start	169
20	byd	391	40	wait	228	60	issue	169

Fig. 2.

Word frequency statistics (word cloud).

The Word Frequency Statistics and word cloud in our study include key terms such as “vehicle”, “price”, “electric vehicle”, “Malaysia”, “China”, “battery”, “affordable”, “expensive”, “government”, “charge time”, “Tesla”, “sell”, and “market”. These terms are central in discussions related to the general concern of EVs in the community. “Vehicle” and “electric vehicle” suggest a focus on the various EV models and their features while “Price” alongside “affordable” and “expensive” are indicative of the financial aspects of EV ownership. More specifically, the latter suggests the cost considerations that consumers face. Additionally, “Battery” and “charging time” relate to the technical concerns of EV technology. “Malaysia” and “China” represent geographical influences where Malaysia is inevitably the focus of national market discussions and China is a leading EV producer. Policy discussions impacting EV adoption are likely referred to in the term “Government” while brand influence and market trends are suggested through the terms “Tesla”, “sell”, and “market”. Using the word cloud as a starting point, we explore the intrinsic connections between these terms, uncovering underlying themes and relationships within the dataset in the next sections.

Market interest topics

Following prior research, topic coherence was used to guide the selection of the number of topics⁷³. Specifically, we evaluated models with 2 to 10 topics, and the optimal number of topics was determined based on the C_v coherence score, computed using the CoherenceModel implemented in the Gensim library⁷⁴, as shown in Fig. 3. Although the coherence score was highest at three topics, we observed that the four-topic model (K = 4) offered substantially better interpretability and topic distinctiveness. At K = 3, key dimensions were conflated; however, increasing the parameter to four effectively disentangled these themes without compromising model stability, as evidenced by the high coherence score (~ 0.49) before the sharp drop-off at K = 5^75,76. Based on this refined separation, we labelled the four major themes as market trends, government support, battery technology, and product attributes (see Figs. 4, 5, 6 and 7). The circles in the left panel of these figures represent different topics where their sizes and quantities correspond to the topic’s importance. The placements of these circles demonstrate the proximity of the relationship between these topics. Alongside these data, a horizontal histogram with 30 bars lists the key terms that are necessary to understand each subject. These key terms allow for the identification of specific market focus. As the LDA model does not confine a word exclusively to one topic, a word may appear in several topic groups¹⁹. The results from the LDA model are then combined with the manual interpretation of relevant news content to analyse consumer’s interest.

Fig. 3.

LDA coherence score.

Fig. 4.

LDA topic model for Paultan.org comments—Topic 1.

Fig. 5.

LDA topic model for Paultan.org comments—Topic 2.

Fig. 6.

LDA topic model for Paultan.org comments—Topic 3.

Fig. 7.

LDA topic model for Paultan.org comments—Topic 4.

Interestingly, the two most popular topics, Topic 1 and Topic 2, are related to the theme “price”, aligning with the findings of Hamzah et al.¹⁸. The authors identified value proposition as one of the main themes. Several studies^77–79 also identified that price is a particularly important factor for EV adoption. Adepetu and Keshav⁷⁷ further underscored this by discovering in Los Angeles that price is even more important to consumers than driving range. Upon closer examination, the specific terms included in these topics reveal two distinct aspects of price that concern the Malaysian consumer: market trends and government support.

Topic 1 captures price–market discourse as a central interpretive frame through which consumers evaluate EV adoption (Fig. 4). Terms such as “selling”, “buying”, and “sales” suggest that Malaysian commenters are not merely discussing EVs as a technology, but as a market-mediated innovation whose uptake depends on affordability and credible price trajectories. This aligns with DOI logic: price is a salient component of perceived relative advantage, and adoption is more likely when an innovation is seen as economically superior to existing alternatives. Recent Malaysia-focused evidence similarly indicates that high upfront prices remain a major barrier to EV diffusion, reinforcing why price-related discourse dominates consumer conversations⁸⁰.

The prominence of both “Malaysia” and “China” indicates that consumers tend to assess Malaysia’s EV market by benchmarking it against more mature and competitive EV markets, particularly China. The clustering of “brand” with manufacturers such as “Tesla”, “BYD”, “Geely”, and “Toyota” implies that adoption expectations are shaped not only by product-level pricing, but also by perceived brand rivalry and market scale, which consumers associate with faster price declines through competition and learning effects. Importantly, repeated references to Thailand and “manufacturer” indicate that Malaysian commenters interpret EV affordability through regional production and investment signals, treating neighboring industrial developments as evidence of feasibility. In DOI terms, this reflects the role of observability: visible outcomes elsewhere (e.g., new EV plants and mass production timelines) become persuasive cues that an innovation is moving toward mainstream affordability.

A notable feature of this topic is the appearance of Malaysia’s national carmakers, “Proton” and “Perodua”. Although they are not yet established EV producers, the government’s optimism that the first national EV can be produced by 2025 positions these firms as symbolic anchors of domestic readiness⁸¹. Consistent with earlier findings that local consumers scrutinize the pace of national manufacturers’ progress in electrification, the salience of Proton/Perodua in this topic indicates that EV adoption is being evaluated as an ecosystem or institutional readiness issue, not solely an individual purchase decision⁸². This resonates with ecosystem-oriented adoption research, which argues that EV diffusion depends on coordinated production capacity, supply chains, policy credibility, and infrastructure—conditions that shape perceived market stability and long-term affordability.

As further evidence, the terms “local”, “Thailand”, and “manufacturer” recur frequently, reflecting discussions of EV investment activity and the establishment of manufacturing facilities in Thailand. The following comments illustrate these concerns:

BYD has gone very far ahead in Thailand now producing the seal and its blade batteries (…) more than seven Chinese companies are also investing, means building manufacturing plants and showrooms all over Thailand.

This is very similar to what Chang An motors Thailand is planning a mass production of cheap EV to the world Thai made cheap EV by 2025

Noticed they mentioned Indonesia and Thailand but not Malaysia? We are no longer the tiger we once were. The sooner we realise the better for us. Then we can start proper restructuring if we are serious in the automotive game.

This contrast helps explain Malaysia’s EV adoption dynamic in two ways. First, the lack of strong local manufacturing may lead to higher EV prices and fewer model choices, since vehicles are mainly imported. This can slow down adoption. Second, the comments reflect a perception that Malaysia is falling behind regional peers. This may reduce consumer confidence in the long-term development of EV infrastructure and policy support.

Topic 2 (Fig. 5) centres on government support (or intervention), heavily emphasising the financial aspects of purchasing EVs with terms like “buy”, “afford”, and “pay”. Several terms associated with “government” are also observed including “road tax”, “tax”, “subsidy”, “incentive”, and “country”. These can be said to reflect the market’s keen interest in understanding how governmental policies and tax laws can ease the financial burden of EV transition.

When it comes to “road tax”, Malaysia’s Road Transport Department (RTD) stipulates that the road tax for traditional ICE vehicles is determined by factors such as vehicle emissions (e.g., 1500 cc) and vehicle type. Many users argued that since EVs produce almost zero emissions, owners should be exempted or be subjected to a minimal amount. Currently, EVs are exempted from road tax until 2025. Moving forward, the road tax for EVs is expected to be higher than for ICE vehicles according to RTD⁸³. This projection generated substantial concern in the public discourse, where users frequently conflated electricity SST, charging costs, and road‐tax revisions. For example, comments such as.

don’t forget about the high (really high) EV road tax after the exemption ends in 2025.

Better to keep my ICE where government pay the petrol subsidy… than buying an EV and pay electricity tax to the government

illustrate widespread uncertainty and misconceptions about the structure and rationale of forthcoming EV road‐tax reforms. Unsurprisingly, subsidies and incentives also gained attention among the Malaysian consumers which several studies have shown to be closely related to widespread EV adoption^84–86. These findings indicate that public sentiment is influenced less by the absolute level of road tax and more by the perceived uncertainty surrounding future policy directions. To address this issue, policymakers would benefit from adopting clearer and more consistent communication practices. This includes providing accessible explanations of how road tax for EVs will be calculated, outlining expected timelines for policy implementation, and presenting comparative estimates that show how EV ownership costs differ from those of ICE vehicles. It is also important to differentiate road-tax reform from other cost components such as electricity tariff changes and charging-infrastructure expenses. Clear distinctions across these elements will help current and prospective EV users develop a more accurate understanding of long-term ownership costs and may reduce the anxiety that arises from policy ambiguity.

The energy price from the Malaysian context adds another layer to this discussion. With the existing subsidies on gasoline, low petrol prices can serve as a barrier to the adoption of EVs. This theme is echoed in,

Same reasons why EV adoption is slow in Malaysia. Because 1. charging [is] more expensive than petrol; 2. tier-up your TNB (electricity company) bill becoming expensive….

Past studies have highlighted that gasoline prices correlated to EV acceptance^42,79, making terms like “fuel” and “petrol” central. This finding reflects the interplay between existing fuel subsidies and the promotion of EVs.

Compared with EV sentiment studies in other national contexts, the Malaysian corpus shows a more explicit distributive justice framing. The frequent appearance of terms such as “T20,” “B40,” and “poor” points to a deeply normative dimension in public sentiment, in which EV policies are evaluated through distributive justice considerations rather than through affordability alone. These references indicate that commenters are not simply assessing whether EVs are financially accessible, but are actively interrogating who is entitled to policy benefits and who ultimately bears the costs of transition. In this sense, cost is not framed as a universal constraint, but as a socially differentiated burden whose legitimacy depends on how benefits and responsibilities are allocated across income groups.

This interpretation becomes clearer when compared with text-mining and sentiment studies from other EV markets. In China, text-mining studies using Weibo and online behavioural data show that EV discourse is often shaped by product attributes, price, powertrain type, regional variation, and public attention^19,87,88 In India, social-media-based sentiment analysis shows that price, maintenance, and safety are among the most frequently discussed issues²⁵. These findings suggest that consumers in emerging EV markets often evaluate EVs through affordability and functional-risk considerations. While the Malaysian corpus also reflects these concerns, it further shows that policy incentives are not interpreted merely as cost-reducing instruments. Rather, they are assessed in relation to perceived economic need, fairness in public support, and the legitimacy of state intervention in the EV transition.

Ruan and Lv’s studies suggest that Western EV discourse is strongly shaped by political, media, and environmental framings, rather than by adoption barriers alone. Their Reddit and Twitter analyses show that discussions vary across online communities and platforms, with political actors and news media playing a visible role in shaping EV-related sentiment and topics^54,55. However, these studies do not report the same explicit income-group dispute found in the Malaysian corpus.

Similar critiques have been raised in other national contexts. In the United States, clean-energy tax credits have been shown to disproportionately benefit wealthier households, prompting debates about the regressive effects of incentive-based climate policies⁸⁹. The Malaysian corpus demonstrates that comparable distributive concerns emerge organically in public discourse, suggesting that EV policies are subject to moral evaluation even in the absence of explicit justice framing by policymakers. This indicates that public acceptance of EV incentives is shaped not only by economic calculus, but also by perceptions of whether policy outcomes align with widely held notions of social fairness.

This interpretation is consistent with Hamzah et al.¹⁸, who found that Malaysian perceptions of EV policy are structured around income-based classifications and distributive concerns. It is further reinforced by recent empirical evidence showing that lower-income consumers are more responsive to EV subsidies when these are explicitly designed to address equity and affordability constraints⁹⁰. Together, these findings suggest that distributive justice operates as a critical mediating mechanism between policy incentives and public acceptance. Where incentives are perceived to favour elites, they may undermine rather than enhance support for EV adoption, regardless of their financial generosity.

More broadly, this pattern situates EV incentives within Malaysia’s wider moral economy of transport subsidies. In a low-fuel-price and subsidy-based transport environment, policy incentives are not interpreted merely as financial tools, but as public benefits whose credibility depends on perceived fairness, stability, and social justification.

The third topic as shown in Fig. 6 highlights consumer attention surrounding EV batteries and power, focusing on technical features such as the battery capacity of EVs. This topic is a pressing concern among consumers when it comes to electric vehicles^91,92. In addition, existing EV charging infrastructures are also discussed including the distribution and strategic placement of charging stations as well as the integration of charging facilities in parking lots. This discussion is crucial in supporting the daily practicality of EVs. In terms of consumer convenience and the overall feasibility of EV adoption, discussions extend to charging time and the availability of different types of chargers.

In addition, discussions on charging infrastructure, including station distribution and integration into parking spaces, indicate concerns about the practical usability of EVs. Limited infrastructure may increase perceived complexity and uncertainty, which can delay adoption. Conversely, improvements in accessibility may enhance trialability and observability, making EV usage more visible and easier to experience.

Viewed through the lens of DOI theory, this topic reflects how consumers evaluate EVs based on their perceived innovation attributes. Concerns over battery capacity, charging time, and charging station availability point to perceived complexity and uncertainty in daily use, while improvements in charging accessibility may enhance trialability and observability by making EV use easier to experience and more visible. The frequent comparison between EVs and ICE vehicles further indicates that consumers are assessing the relative advantage and compatibility of EVs against familiar gasoline-powered alternatives. Thus, the repeated references to fuel, gasoline, and diesel suggest that Malaysian consumers are still negotiating the transition from conventional vehicles to EVs, positioning EV adoption largely within the persuasion stage of the diffusion process.

The fourth and final topic (Fig. 7) focuses on the essential product attributes of EVs, accounting for the technical and aesthetic aspects of automotive design. This topic includes components and features like “engine”, “design”, “interior”, “turbo”, and “rear” which highlights EV technological and design innovations. Terms like “turbo” and “interior” emphasise growing consumer interest in both performance capabilities and comfort features. This somewhat reflects consumers’ appreciation for the engineering and ergonomic features of EVs.

Interestingly, the discussion is also observed to include branding and market positioning of well-known automobile manufacturers such as Mercedes, BMW and Volvo. The frequent mention of these brands suggests that the manufacturer’s reputation and brand image play an important role in their EV purchasing decision for the Malaysian consumer⁹³. Beyond functional attributes, brand identity associated with luxury, reliability, and after-sales service serves as an important signal of product quality. Recent studies show that strong brand reputation can reduce perceived risk and increase consumer trust, particularly in the context of emerging technologies such as EVs^94,95.

This pattern may partly reflect the fact that EV adoption in many developing countries remains at a relatively early stage of diffusion⁸². At this stage, public familiarity with EV technology is often still limited, while uncertainty regarding battery durability, charging access, maintenance, and long-term value remains relatively high. Under such conditions, consumers are less likely to rely solely on direct knowledge or accumulated experience, and more likely to depend on external cues when evaluating EVs. Brand reputation therefore becomes especially important, as it can signal reliability, reduce perceived risk, and make a relatively unfamiliar technology easier to assess. In this sense, well-known automobile brands may help reduce the perceived complexity of EV adoption and increase the observability of EV technology by making it more familiar, visible, and credible to potential consumers.

Taken together, the four LDA topics show that online discussion of EV adoption is shaped by several key dimensions, namely market trends, government incentives, battery and charging, and product features. Table 2 presents a summary of the four LDA topics together with their most prevalent keywords, core themes, and representative anonymised comments.

Table 2.

Summary of the four LDA topics and representative comments.

LDA topic and prevalence	Most prevalent keywords	Core theme	Representative anonymised user quotes
Market trends (31.3%)	Price, Malaysia, China, Tesla, BYD, Proton, brand, market, model, buy	Focuses on EV market development in Malaysia, especially pricing, brand competition, and consumer interest in major EV players	It (EVs) is still too expensive for one and range (real world) is about 25% off what would cover 95% of people. But EV tech won’t stay at 2024 levels forever. Production efficiencies, especially with battery tech (density and hence weight/range, charging speed, price) will lead to much lower prices. In that way, EV sales is not going to pick up. paying more for poor charging infrastructure while u pay less and get the convenience of ICE engine.
Government incentives (27.2%)	Government, tax, road-tax, subsidy, incentive, affordable, expensive, cost, pay, buy	Reflects public concern about whether policy support can make EVs more affordable and easier to adopt	The government should increase their investment in chargers in order to have more control over the pricing which eventually is the key determinant in whether the adoption of EV will be successful in the future.
Battery and charging (22.4%)	Battery, charge, range, charger, charging, charge-time, station, power, electricity, fire	Captures concerns about battery performance, charging convenience, infrastructure, and safety	Should have more DC chargers as well especially at highways so that have no range anxiety. 7kw ac is wall charging is just too slow, take 15 h for full charge
Product features (18.6%)	Mercedes, BMW, design, interior, drive, SUV, model, seat, light, brand	Highlights discussion of vehicle design, driving experience, comfort, and brand image	The most beautiful brand new [Brand name] on sale right now. ANGEL EYES OVER SQUINT EYES! The interior looks over-design and too much stuffs going on.

Market sentiments analysis

Subsequently, the VADER tool is utilised to carry out a sentiment analysis of the comments. Sentiments are classified using a numerical score ranging from -1 to 1 for each text. A score of 0 indicates no positive or negative emotional content and is considered neutral. Scores nearer to 1 indicate increasingly positive sentiments, while scores nearer to − 1 are considered increasingly negative. Table 3 explains this scoring mechanism in relation to the specific analytical results of this study. As a post hoc validation step, we randomly sampled 300 comments and compared VADER-based sentiment labels with human annotations. The manual coding scheme used three discrete sentiment categories: positive (1), neutral (0), and negative (-1), rather than continuous sentiment values. To ensure comparability, VADER outputs were mapped onto the same three-category scale prior to evaluation. Model performance was then assessed using the F1 score, which is a standard classification metric that combines precision and recall to provide a balanced measure of predictive accuracy across categories⁹⁶. In this study, both micro F1 and macro F1 scores were reported: micro F1 reflects the overall classification performance across all samples, while macro F1 gives equal weight to each sentiment class by averaging their individual F1 scores. Using this scheme, the model achieved a micro F1 score of 0.677 and a macro F1 score of 0.675. The closeness between these two scores indicates relatively balanced performance across classes. At the class level, VADER performed best on negative comments (F1 = 0.705), followed by positive comments (F1 = 0.694), while performance on neutral comments was lower (F1 = 0.624), suggesting that neutral sentiments were comparatively more difficult to classify accurately. Based on this classification, the results are compiled into time series statistics where the overall average sentiment score, as defined by VADER, is 0.12 (See Fig. 8). This score indicates a distribution of 48% positive comments (scores greater than 0), 25% neutral comments (scores equal to 0), and 27% negative comments (scores less than 0). Generally, Malaysians are then observed to have mildly positive sentiments towards EV.

Table 3.

Examples of VADER sentiment analysis results from comments.

Comments (excerpted)	Sentiment scores	Sentiment orientation
I think they should consider Malaysia. We have expertise in making battery. And we have graphene producer in Malaysia, (…) (EV Brand name) sodium battery can be manufacture here. What say you (EV Brand name), come to Malaysia	0.34	Positive
The government should encourage more higher quality investment in EV and green impact like (Brand name). We should be a part of the supply chain in reducing carbon emission, (…) Kudos to (Brand name) for choosing Malaysia	0.8591	Positive
How much will it cost to change Lithium battery packs 77.4 kWh after 6–8 years warranty?	0	Neutral
Will existing EV owners be able to switch to the new RFID EV plates?	0	Neutral
Other countries forbid the charging and sometimes even parking of EV’s at basement level too	− 0.3182	Negative
I don’t understand why government is pushing so hard on EV. By right we should be pushing for hydrogen. Sarawak is forward thinking in this. We are relying heavily on coal and dam for power. (…)	− 0.1447	Negative

Fig. 8.

Sentiment distribution of comments on Paultan.org.

Our time series analysis (Fig. 9) demonstrates that despite fluctuations in discussion intensity (number of comments), the Malaysian market sentiment towards EV has consistently maintained a slight positive trend (0.1 to 0.15) over the course of 14 months.

Fig. 9.

VADER time series results.

To this end, the Malaysian market sentiment towards EVs is generally neutral to positive (or “cautiously optimistic”), contrasting the findings of Wu et al.¹⁹ in China. The authors reported a strong positive sentiment towards EVs. On the other hand, Jena²⁵ observed predominantly negative sentiments in India.

We argue that Malaysian consumers’ “cautiously optimistic” sentiment toward electric vehicles (EVs) reflects a hopeful yet conditional orientation that is shaped by both perceived benefits and unresolved uncertainties. This sentiment can be meaningfully interpreted through the lens of Rogers’ Diffusion of Innovation theory, which suggests that adoption decisions depend not only on the perceived relative advantage of an innovation, but also on its compatibility with existing practices, perceived complexity, and the clarity of its long-term outcomes.

To understand the market’s hesitancy, one must consider the sheer scale of Malaysia’s established car culture. Ranking as Asia’s second most vehicle-dense nation⁹⁷, Malaysia’s reliance on private transport creates a high “switching cost” for consumers. While the environmental benefits of EVs drive optimism, the practical risks of disrupting a highly functional, car-centric lifestyle fuel the caution. This sentiment reflects a market that is intellectually ready for innovation, but structurally tethered to tradition. This entrenched car culture creates a psychological paradox: while Malaysians are quick to recognize the relative advantages of EVs in terms of sustainability, they are equally sensitive to any potential disruption to their seamless ICE-based mobility. For these consumers, the hurdle isn’t just embracing new technology; it is ensuring the innovation can achieve full compatibility with a lifestyle built entirely around the convenience of fossil fuels.

In diffusion terms, optimism arises from the growing awareness of EV benefits and increasing policy signals promoting sustainable transport. However, caution persists due to perceived complexity and uncertainty surrounding affordability, charging infrastructure readiness, battery durability, and maintenance costs. In addition, skepticism regarding the consistency and long-term credibility of policy support may limit the observability of successful EV adoption outcomes, further reinforcing adoption hesitancy. These concerns suggest that while consumers are open to innovation, they require stronger institutional support and clearer evidence before making a sustained commitment.

Our findings indicate that although the growth trajectory of the Malaysian EV market appears promising, its success ultimately depends on how effectively government and industry stakeholders address these diffusion barriers. Consumer knowledge plays a critical role in mitigating perceived technical and economic constraints⁹⁸. As such, a strategic approach emphasizing transparent communication, public education, and tangible improvements in affordability and infrastructure accessibility is essential. By balancing optimism with a pragmatic response to these structural concerns, policymakers and manufacturers can facilitate a smoother transition that aligns technological advancement with consumer readiness and acceptance, thereby supporting Malaysia’s broader shift toward a more sustainable transport system.

Conclusion

Electric vehicle adoption in Malaysia remains at an early stage, particularly as EVs are not yet widely perceived as financially advantageous when compared with conventional internal combustion engine vehicles⁴¹. As a developing sustainable technology, close attention to market sentiment is important because public perceptions may shape the pace and direction of adoption. Drawing on comments collected from Malaysian automotive news platforms, this study examined consumer sentiment and attitudes towards EVs through text mining techniques. The findings show that Malaysian consumers most frequently discuss price and market trends, followed by government support, battery and charging issues, and product features.

Beyond identifying these core discussion themes, the study also shows the value of using unstructured online discourse to examine EV sentiment. The analysis revealed patterns that would likely be less visible in survey-based designs, including misconceptions about EV policies such as subsidy eligibility and charging infrastructure support, as well as concerns about distributive fairness, where EV adoption is seen by some users as favouring higher-income groups while shifting indirect costs to others. These perceptions are often expressed through informal narratives, emotional language, and peer-to-peer interaction rather than direct evaluative statements. Such forms of expression are often reduced in survey responses because of social desirability bias and the limits of predefined response options⁹⁹. In this way, online comment data provide a more natural space for latent perceptions and contested meanings to surface.

Overall, the findings suggest that market sentiment towards EVs in Malaysia is shaped not only by practical issues such as price, policy, and technical performance, but also by misunderstanding, trust, and perceived fairness. These insights point to the need for more targeted policy communication and strategic interventions to support EV adoption, in line with Malaysia’s broader climate commitments and Sustainable Development Goal 13 on Climate Action.

Based on the findings and discussion presented in this study, the following policy suggestions are proposed for promoting EV adoption in Malaysia:

Encourage broader EV model availability to address affordability concerns

Consumer attention to pricing, affordability, and brand variation suggests that the current market lacks sufficient diversity, particularly in the entry and mid-level segments.In the Malaysian context, this concern often appeared in the form of comparisons with Proton and Perodua, indicating that EV prices are evaluated against familiar national ICE benchmarks rather than against other EVs alone.

Policy measures should focus on facilitating the entry of competitively priced foreign models while also incentivizing manufacturers to offer a wider spectrum of options. Strategic tariff adjustments, streamlined type approvals, and time-bound fiscal incentives may help increase supply elasticity and trigger more responsive pricing structures.

Strengthen domestic EV production capacity to enhance national competitiveness and consumer confidence

Mentions of national brands and comparisons with regional counterparts reflect concern over Malaysia’s positioning in the emerging EV economy. Such comments suggest an expectation that Malaysia should not remain a passive EV importer, given its long-standing automotive manufacturing base and national car identity. To address this, industrial policy should prioritize the local development of EVs through targeted investment in R&D, tax incentives for domestic production, and public–private partnerships. A robust local EV sector would not only foster economic spillovers but also signal long-term commitment to electrification, reducing public hesitation around technological transition.

Recalibrate fiscal incentives to ensure clarity, accessibility, and socioeconomic fairness

Conversations around subsidies, road tax, and income brackets point to a need for greater transparency and equity in policy design. Notably, frequent references to B40, M40, and T20 in user comments indicate that Malaysians frame EV incentives in terms of distributive fairness rather than purely environmental benefit. The government should consider refining existing incentive schemes by embedding clearer eligibility criteria, reducing administrative barriers, and ensuring that benefits are distributed progressively. Particular attention should be paid to the B40 and M40 groups, ensuring that adoption is not limited to affluent segments.

Expand public education and information dissemination on EV ownership

Persistent uncertainties about battery life, replacement costs, charging expenses, and long-term maintenance continue to create confusion among potential EV users. These knowledge gaps limit public confidence and slow adoption. One way to address this issue is through public education campaigns using digital media, automotive exhibitions, and community outreach. These efforts should help people understand what owning an EV actually involves and how it compares to traditional ICE vehicles, both in terms of cost and environmental impact. As highlighted by Basri et al.¹⁰⁰, reducing these misconceptions through education is especially important in countries like Malaysia, where public understanding of EV ownership remains limited.

At the same time, our Topic 2 findings show that much of the public’s confusion stems not only from technical uncertainties but also from unclear government communication regarding road-tax reforms, subsidies, and incentives. Users frequently conflated electricity tariffs, charging costs, and future EV road-tax structures, leading to heightened anxiety about long-term affordability. The repeated appearance of terms such as “road tax,” “subsidy,” “incentive,” and income-group markers like “T20” and “B40” further indicates that Malaysians interpret EV policies through a lens of financial burden and distributive justice. To address these intertwined concerns, public-facing education efforts should explicitly clarify how EV-related policies work, how road tax will be calculated once exemptions end, and how incentives compare across income groups. In this way, education campaigns would not only correct misconceptions about EV technology, but also reduce policy uncertainty-a problem that our corpus shows is equally important in shaping consumer perceptions.

Accelerate the development of accessible and reliable charging infrastructure

Discussion reveals that concerns about charging station availability and practicality remain significant barriers. National infrastructure strategies should therefore ensure wide spatial coverage, including semi-urban and rural areas, and support interoperability between networks. Further, policies should incentivize both private-sector investment and local-level planning, with emphasis on long-term operational sustainability rather than short-term deployment numbers.

Limitations and directions for future research

While this study contributes to the growing interest in EV adoption, sentiment research in the Malaysian context remains limited. By applying text mining to user-generated content, this work is explicitly exploratory in nature and aims to generate preliminary insights and hypotheses rather than produce generalizable population-level estimates. The findings should therefore be interpreted as an initial mapping of salient themes and concerns in Malaysian online discourse on EVs, which can inform subsequent theory-driven and empirically validated research using complementary methods.

Several limitations warrant a more critical reflection, particularly the reliance on Paultan.org, whose audience is generally tech-savvy and male-skewed. While the findings offer valuable insights, this demographic profile likely creates an “enthusiast bias,” where sentiment patterns overemphasize technical performance, torque, and gadgetry relative to broader everyday mobility needs. More importantly, this sampling structure may shape not only what concerns are expressed, but also how they are evaluated, suggesting that the observed sentiment patterns are more representative of early adopters rather than the general population. For instance, the identified “cautious optimism” may reflect a technophile’s willingness to tolerate infrastructure gaps and policy uncertainty in exchange for early access to innovation, whereas mainstream users may place greater emphasis on practical considerations such as charging accessibility, cost stability, and household usability. As a result, the relatively stable positive sentiment observed in this study may overestimate overall market readiness if interpreted as representative of the broader population. Instead, it should be understood as reflecting a segment that is inherently more receptive to technological novelty, consistent with early-adopter behaviour in the Diffusion of Innovation framework. Consequently, the findings should be interpreted as a leading indicator of early market sentiment rather than a universal consensus. Future research incorporating more diverse demographic, linguistic, and regional groups is necessary to assess whether these patterns persist across the wider Malaysian population and across later stages of technology diffusion.

Second, the study focused primarily on English-language comments. Although Paultan.org’s editorial content is predominantly in English and the majority of user contributions also appear in English, comments in other languages (e.g., Malay, Mandarin, Tamil) were excluded. These accounted for less than 10% of the total dataset and were often informal or code-switched. Nevertheless, Malaysia’s multilingual environment implies that linguistic filtering may omit nuances in sentiment expression across different language groups. Future studies could address this by incorporating multilingual or code-switched data to achieve broader representativeness.

Third, the limited availability of EV-related data in Malaysia, such as detailed sales figures, demographic breakdowns of adopters, and regional adoption patterns, restricted the potential for deeper empirical analysis including triangulation, cross-validation, and comparative studies.

Despite these limitations, the study provides meaningful insights into the attitudes of a specific, engaged consumer segment and highlights key issues such as affordability and government trust. Future research can build on this work by incorporating multilingual data, improving sentiment tools, and combining text mining with other methodologies to better capture the full spectrum of public sentiment in Malaysia.

Appendix 1

Text preprocessing rules and sample cleaned comment

See Tables 4 and 5.

Table 4.

Summary of preprocessing rules.

Preprocessing procedure	Description	Example
Text normalisation	Lowercasing, punctuation removal, stop-word removal, tokenisation, and lemmatisation were applied using NLTK	“buyers” → ”buyer”
Numerical expression transformation	Selected numerical expressions were transformed into meaningful standardised terms using regular expressions	“RM150k” → ”price”; “50 kWh” → ”battery-capacity”
Thesaurus-based standardisation	Abbreviations, local slang, near-synonyms, and selected vehicle-related terms were standardised	“govt” → ”government”; “cheap” → ”affordable”; “ICE cars” → ”ICE”
Compound-term standardisation	Frequently adjacent words were joined using hyphens to preserve phrase-level meaning	“road tax” → ”road-tax”
Language filtering	Non-English comments were removed using Langid	Non-English text → removed

Table 5.

Example of raw and cleaned comment.

Raw comment	Cleaned comment
The new EV looks interesting, but RM150k is still too expensive for buyers compared with ICE cars	New ev look interesting price expensive buyer compare ice

Author contributions

D.Q.: methodology, data curation, investigation; Formal analysis, writing—original draft. LC: conceptualization, funding acquisition, resources, supervision, investigation, writing—review and editing. A.R.A.S: funding acquisition, supervision, writing —review and editing.

Funding

This work was funded by the Ministry of Higher Education Malaysia under Fundamental Research Grant Scheme (FRGS/1/2023/SS06/UPM/02/1).

Data availability

The raw/processed dataset and related Python scripts used in this study are currently part of an ongoing research project and are therefore not publicly available at this stage.

Declarations

Competing interests

The authors declare no competing interests.