Application of artificial intelligence based on the fuzzy control algorithm in enterprise innovation

Abstract

Artificial Intelligence (AI) has gained immense popularity in recent years as many enterprises have realized their potential to change the way of conducting business innovatively. The new concepts, items, or procedures are developed and implemented within a business or organization to enhance productivity, effectiveness, and competitiveness, and this is called Enterprise Innovation (EI). AI techniques are required to make decisions more effectively in challenging and dynamic situations, like EI, as result of competitive marketplace. Hence, an intelligent, innovative strategy with Q-learning and Takagi Sugeno Fuzzy Control (Q-TSFC) algorithm has been proposed as it combines adaptive learning and Fuzzy Logic (FL) that humans understand to improve decision-making in enterprise innovation. Q-learning seeks to maximize the enterprise's profit by utilizing the newly acquired knowledge, exploring activities, and adaptive learning based on the optimal ε greedy policy that results with rewards and the experiences. To develop a framework for making decisions and connections between input from the learned Q values and output decisions using enterprise expertise and linguistic conventions. The objective is to handle language uncertainty and imprecision in market trend. So, it leads to right decisions even without accurate numerical facts. The proposed approach is validated by evaluating metrics like cost savings, customer satisfaction, and innovation performance efficiency in the competitive edge in the market. With this proposed Q-TSFC algorithm, the obtained results are 96.5% customer satisfaction ratio, 96% enterprise performance efficiency, cost savings of about 48% profitable value, and the coefficient of determination R² is 0.83, respectively.

1. Introduction

Enterprise is moving into a new phase of intense innovation and quick expansion with the capacity to develop new forms of collaboration which is made possible by the expanding enterprise evolution and the newest AI. The enterprise may involve companies, quasi-corporations, organizations, and unstructured businesses like partnerships and sole proprietorships. Automation of repetitive processes is to enhance the consumer experiences and the promotion of innovation and growth. They are all made possible by AI and also through the impact of the enterprises for greater benefit in some ways. The data from multiple input forms are fed simultaneously, especially when precise input information is unnecessary; fuzzy systems offer greater resistance against erroneous, distorted, or incomplete input data. Fuzzy control algorithms based on AI can be applied in various ways to business innovation. It is possible to utilize Fuzzy control to solve complicated and confusing real-world situations because it models unpredictable and inaccurate inputs using Fuzzy logic.

For EI, where unpredictability is common, Fuzzy control is a good fit because it can successfully represent and manage ambiguity and imprecision in innovation decision-making through linguistic conventions and Fuzzy membership functions. It has been shown that Fuzzy modeling works effectively when technical procedures have become too complicated to analyze using commonly utilized quantitative approaches or when information sources are available in the enterprise but are interpreted incorrectly and also repeatedly. By analyzing market trends, consumer behavior, and competitor data, AI-powered systems assist enterprises in designing and creating innovative products. The application of AI helps enterprises to produce unique new products which are better and that meets the demands of their clients. AI is assisting enterprises in decreasing costs and boosting profitability by optimizing the product phase of development and shortening the time to market. Using AI algorithms, it is urgently necessary to analyze the control effects to advise enterprises on the best course of action. Fuzzy mathematical techniques are employed in systems for expertise and AI since many criteria and regulations cannot be precisely defined. The answers provided by the fuzzy approaches are more accurate if the systems have a significant amount of information and associated experience that is relatively intricate and large.

An organization that incorporates pioneering procedures, solutions, or merchandise is said to be engaging in EI. So, inventive thinking in business is crucial for an enterprise to thrive in today's uncertain market [1,2]. The analysis of the numerous effects of AI focuses on both the good and negative effects on organizations, businesses, communities, and people, and [3] find that the significant academic developments and innovations in the enterprises, as well as its impact on business ventures and, ultimately, the worldwide market. An enterprise's innovative business approach is the foundation for its management and the guarantee of accomplishing goals for its ongoing operations and potential growth based on the different innovations. The key ingredients for developing an enterprise's successful model for business were identified, and the components for thriving business and its strategy that should be presented for the greatest efficacy were investigated [4]. To make a successful enterprise with financial innovation ideas [5] investigated the association between innovation and the ability of small and medium-sized enterprises to access outside finance. All sorts of innovations, particularly the debut of new goods and technology with the advancement of the current ones, continue to be strongly and favourably connected with long-term financial obligations. The EI information can often be vague and include quantitative and qualitative components [6]. Hence, these reasons make it essential to include subjectivity, variability, and imprecision in conventional decision-making systems. To meet the need for real-time decision-making, it is necessary to enhance that decision-making process in a way that is not usual for an environment that will allow an artificial intelligence platform to replace employees. The relationship between AI and the management of EI [7] was examined, and the results showed that AI reshapes the businesses, controls innovation handling within them, and compels management in the investigated companies to reconsider the process of innovation from an organizational perspective.It is based upon the data stored in the reasoning system's knowledge base, and it presents all the possibilities between agreement and disagreement. It mimics on the way humans make their decisions, especially when faced with uncertainty. Particularly, in the enterprise sectors like financial industries, these systems are utilized with expert systems and Neural Networks (NNs) with Fuzzy Logic (FL) for anticipating projected profits on the equities, capital portfolio control, and cash flow planning [8,9].In Ref. [10], the professional experience of experts and the investment of EI have a positive relation using the multiple regression analysis with the big data model of small and medium-sized enterprises with an average coefficient of determination of 0.198. Over 70 percent of enterprises employ innovations to increase output to target a certain market. Nearly half of the innovation updated enterprises are motivated by minimialized production costs and also by the application of the neuro FL inference [11]. Modifications in the enterprise environment will eventually impact several financial aspects through their connection to the business's plans, leading to financial risks. The industrial environment's uncertain factors [12] must first be identified for the enterprise's financial notification and identified using a Fuzzy control mechanism to be effective. The comparative analysis of the potential for growth in entrepreneurship using the applied AI techniques and the Fuzzy indicators of the enterprise activity and the principles of AI that are applied for tackling issues with entrepreneurship development and generating business judgments in uncertain circumstances [13]. To make a perfect decision in companies [14], applied AI techniques based on the FL multicriteria approach to address the several issues with traditional ways of supporting decisions to invest in company assets also substantiates capital decisions in items bought on the open market. Market revenue forecasting and product error detection based on AI and Fuzzy Logic algorithms can help enterprises understand the market's future trajectory and the enterprise's inventory [15]; it can be maintained in an acceptable range, if predictions are accurate and optimize their financial returns. Additional innovations and systems for making decisions in an enterprise are modified for selecting the optimal levitation controller in an enterprise using the 3-grade Fuzzy-based multicriteria strategy [16]. However, the result shows varied results for control parameters. For analyzing the control parameters [17], outlined a novel method built around the design of a Fuzzy controller for analyzing the actions of within-company innovation in businesses that enables users to understand the complex nature of the occurrences.

The combined application of Q-learning and Takagi-Sugeno Fuzzy Control (Q-TSFC) provides a strong framework for AI-driven decision-making and strategy optimization in Enterprise Innovation. Due to this combination, the contributions of the proposed algorithm are the following: 1) to predict the best enterprise the innovation strategies Q-learning, learn from experiences, and adapt to change the surroundings and through the optimal policy rewards. 2) to establish a framework that gives linguistic rules for decisions and improve the innovation performance control by TSFC by making the procedure for decisions clearer and more understandable for the stakeholders. 3) To optimize the resource distribution, resulting in increased productivity, customer satisfaction, and innovation performance, eventually boosting the enterprise's competitiveness.

The research paper is divided into the following: Section 2 gives a detailed literature study on the innovation implemented in enterprises analyzed with the application of the various techniques and its challenges. Section 3 describes the proposed Q-TSFC to analyze the application of intelligent, innovative strategies related to EI with various aspects. Section 4 gives the results and discussion with the dataset overview implemented in this research and the numerical performance outcomes of the various performance metrics. Section 5 discusses the conclusion of the implemented approach with results and gives the research direction toward the future.

2. Literature survey

Daksa et al. [18] have discussed that Enterprise Innovation (EI) had attracted the attention of academics and development officials as the foundation for economic growth and explored the sources of innovation by considering firms of various sizes all at once. Policymakers can better concentrate on developing their strategies with the help of the Multivariate Probit (MvP-EI) approach in various sizes and simultaneously. Findings reveal that factors influencing EI across all sizes include web page control, the proportion of permanent workers with higher education, the accessibility of hands-on instruction, and the participation in R&D endeavors. A limitation is that it gives the least performance for methods that estimate using only one equation.

Han and Qian [19] have investigated the influence of COVID-19 on the capacity for innovation of China's listed enterprises including small and the large scaled productions using a Fixed Effect Model. The effect of the Hausman(H) has been tested on the EI over 3000 industrial categories like Information Technology (IT), medicine, real estate, telecom, finance, and other sectors with additional support of FEM. The implementation outcome of the panel data source could increase market confidence and leads to the economic recovery and policy development in the Research and Development (R&D). The regression coefficient results of all variables for given enterprises that showed the R² value is 0.21, with the significant effect of the pandemic as having different impacts for all industries.

Nguyen et al. [20] have researched the impact of innovative activities on credit availability using the Panel Regression Model (PRM) for novel items, improved goods, R&D expenditure, new equipments and machinery, and cutting-edge technology. The data source includes more than 4000 observations from industries comprising micro, small, and medium-sized enterprises with no. of employees and total revenue of the limited databases of the industries like agriculture, forest, fishery, trade and service, industry, and construction-related questionnaires with a labor account. The variation inflation factor and Pearson correlation coefficients are employed to check for multi-collinearity occurrence. As a result, the implementation logistic regression model is utilized to analyze the financial returns and the credit access.

Al-Hawamdeh and Alshaer et al. [21] have discussed that an organizational innovation was impacted by the applications of (AI-ENFL) in commercial banks with the Expert structures(E), Neural network platforms(N), and systems with Fuzzy Logic (FL) variables were evaluated. The research considered innovation in products, processes, and management for the organizational innovation factor, with this application of AI based on FL providing a portion of the funds to the development of the technical infrastructure and applications, as well as striking a balance between the use of technology and the information security hazards to preserve the client privacy. The results showed that applications of AI with an FL system had a considerable impact with Cronbach's alpha coefficient of 0.91 for management innovation with an R² value of a maximum of 0.738 and R² of 0.545 for innovation of products. Here, the sample questionnaire is limited to a small data size of commercial banks.

Leung et al. [22] have shown that enterprise innovation can be regarded as creating and applying the novel FL-based ML method for predictive analytics on huge transportation data. Transportation organizations and public transit companies can optimize their operations and timetables and increase consumer satisfaction and service reliability by precisely predicting transportation waits. For transport businesses and agencies, cost savings may result from cutting delays and maximizing resources.

Bag et al. [23] have investigated how Big(B) data-powered AI(B-AI) affects the generation of consumer, customer, and foreign market intelligence to understand deeply about the ways it impacts Business to Business (B2B)marketing as it is based on the reason that affects the enterprise performance. According to the findings, generating customer, user, and foreign knowledge and AI powered by big data are important.A common method for boosting a firm's standing in the marketplace is knowledge management. The result showed that the mean R² is 0.530, the overall performance of the enterprise with innovation is 0.76%, and the research gap related to consumer information privacy and the lack of handling uncertainties in the information.

Zhong et al. [24] have offered an enterprise innovative approach using reinforcement learning called the Q network technique combined with Extreme Gradient Boosting (QXGB) based AI method to choose the suppliers and foresee their upcoming demand for product manufacture.Construct productive data complexities in management for supply chain systems and provide faster response rates and performance results that are more accurate, productive, and dependable when controlling customer demand in unpredictable real-time scenarios. Using metrics like Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), a precise accuracy of 0.96% of production information and logistics features are provided.

Shakeel et al. [25] have discussed the enterprise's innovative process of addressing an intelligent transportation system by improving the accuracy and reliability by automating the background detection and removing the shadow using both the Gradient Partial Equation and the Robust Principle Fuzzy subtraction (GPE-RPF) approach.These techniques examine every pixel contained in the frame of the video visuals, the shadows cast, and the background in different changes of the illumination conditions. The result showed 0.95% correctness accuracy in addition to the online version of completeness and accuracy.

Hu et al. [26] have discussed the innovation performance of enterprises in China, the external aspects of firms, such as the innovation network, tax system, and the innovation circumstances, as well as the inside variables of the capital, expertise, workers, and the business management, R&D activities with patent rights all the influence enterprise innovation performance. Also, focused on technical factors, capital and the employee workforce, organizational, management, and other externals. The innovation performance improves the entrepreneurship and motivation of start-ups towards the market trends.This study has the disadvantage of concentrating on the study's results of the domestic innovation efficiency and leaving out the research findings of foreign company innovation performance.

Rjoub et al. [27] have proposed an Adaptive Neuro-Fuzzy-assisted K-nearest neighbors' Algorithm (ANF-KNNA) for resolving transition challenges in the finance enterprise sector. A rolling window-based autoregressive model with a chaotic enhanced foraging optimization model analyzed the growth.Advancements when demand changes and new financial tools and innovations permeate the financial sector.The proposed idea gives result with an accuracy of 0.91%, privacy of 90%, and robustness of 0.96%. The study was limited to the banking and financial sectors, where the input samples were limited.

Bahoo et al. [28] have reviewed the AI intersection with a Supply Chain Management (SCM-AI) innovation model, including business models related to products and structure. The market performance related to cloud computing,and the Internet of Things (IoT) has deeply driven the business towards innovation to increase numbers as a result of industrialization, synchronizing of technology, and rapid technical advancement.Businesses have learned to use AI technologies with their innovation procedures to improve competencies and gain an edge over their competitors within limited procedures.

ARSLANKAYA [29] has combined Fuzzy Logic and ANF inference systems for identifying the renovation and risk identification of employees in a courier enterprise. The five factors such as absenteeism, cost, and work done were analyzed for research.The input variable was transformed by Fuzzy Logic and verbal expressions with an expert preference for business innovation. The presented study performance was evaluated using a Root Mean Square Error (RMSE) was 5.26.They attributed the absence of statistical techniques to the incapability to assess the variables that were unable to yield quantitative information.

He et al. [30] have evolved a global productivity index for evaluating green business innovations and assisting distance function using the Fuzzy C Means algorithm (FCMA). The critical conditions represent the enterprise regulations, distinct sub-sectors,and the most significant enterprise scale.A business decision-making element is represented by the set of environmental technology possibilities for production, including intended and unacceptable outputs, for each business sector. The study sample does not cover all industrial sub-sectors because of the availability of data.

The review of the literature explains research works related to enterprise innovations with the application of the various AI techniques, expert systems, Neural Networks, statistical measures, and Fuzzy-based control approaches for the performance evaluation of implemented innovation criteria. Out of these, all advantages mentioned in the literature still lack enterprise innovation applications. However, the existing systems are having difficulties while making decisions during the large data analysis. Hence, the suggested Q-TSFC method is validated in contrast to existing techniques, including AI-ENFL [21], B-AI [23], and QXGB [24] for comparison of EI in various industrial sectors with its performance metrics. The abbreviation utilized in this work is listed in Table 1.

Table 1.

List of abbreviations.

Acronym	Full Form
AI	Artificial Intelligence
EI	Enterprise Innovation
Q-TSFC	Q-learning and Takagi Sugeno Fuzzy Control
FL	Fuzzy Logic
NNs	Neural Networks
QXGB	Q network technique combined with Extreme Gradient Boosting
MvP-EI	Multivariate Probit using Enterprise Innovation
H	Hausman
FEM	Fixed Effect Model
IT	Information Technology
R&D	Research and Development
PRM	Panel Regression Model
AI-ENFL	Artificial Intelligence based Expert structures Neural network platforms Fuzzy Logic
B-AI	Big(B) data-powered AI
QXGB	Q network technique combined with Extreme Gradient Boosting
MAE	Mean Absolute Error
RMSE	Root Mean Square Error Rate
GPE-RPF	Gradient Partial Equation and the Robust Principle of Fuzzy subtraction
ANF-KNNA	Adaptive Neuro-Fuzzy-assisted K-Nearest Neighbors' Algorithm
SCM	Supply Chain Management
FCMA	Fuzzy C Means algorithm
FIS	Fuzzy Inference System
PI	Process Innovation
CRM	Customer Relationship Management

3. Proposed scheme

EI comprises an enhanced product or process or an amalgamation of both made accessible to prospective consumers for products or goods the unit implements for processes that differ considerably from its prior products or processes. New or improved features are implemented after the release of the product. Reinforcement learning is an aspect of machine learning—a discipline within the application of AI; a technique called Q-learning is implemented with Fuzzy control concepts. When new organizational, marketing, or operational procedures are deployed when used in the business's activities, the Fuzzy control algorithm aids in understanding and implementing these learned policies leads to more flexible and intuitive practice than the Q-learning method.This determines the best policy for making decisions in the specified innovation domain in an enterprise. The merging of AI-driven applications through Q-learning and expert knowledge recorded in the Fuzzy rules is possible using TSFC combined with Q-learning. Businesses can make better decisions by combining domain knowledge and historical data through this connection. The AI component that operates in the environment, known as the agent, iteratively learns about the surroundings and independently generates innovation success predictions with the uncertain outcomes of the market attributes using the TSFC algorithm. Fig. 1 shows the proposed work's overall process and is explained below.

Fig. 1.

The overall process of the proposed Q-TSFC algorithm.

With the use of Q-learning, the organization can find a balance between exploitation that is using the Innovation successful methods and exploration called attempting out innovation techniques. The organization is guided by Q-learning to concentrate on the more productive innovation streams while investigating the new options by repeatedly revising Q-values following the experienced benefits. Q-learning aids in determining the ideal order of operations or innovation activities that produce the best outcomes. The organization can make educated judgments and assign funding to innovative decisions with the greatest predicted rewards by choosing activities with higher Q-values. The enterprise innovation environment and input conditions can change frequently and are determined using Fuzzy control. Fuzzy control may adapt innovation methods to change the market dynamics and client preferences in enterprise innovation, rendering it less rigid and adaptable. Fuzzy control enables rule-based decisions based on the knowledge of subject-matter specialists like executive managers of enterprises or enterprise unit heads. The organization can easily incorporate human insights into innovation since the control rules may collect expert knowledge.

3.1. Innovative strategies

The main category of EI performance-related criteria to maintain economic growth and analyze the organization's profit using the strategies is listed in Fig. 2. The investment in EI refers to the labor expenses and the capital outlays associated with innovation in processes, product or service Innovation, organizational change, and marketing innovation. There are various indicators used to analyze the EI and the general innovation ratio concerning the term product is given in Eq (1) and the indicators are described below:

$$I_R=\frac{No.ofI}{Totalproducts}*100$$ (1)

Fig. 2.

Innovation strategy pipeline for enterprise performance improvement.

3.1.1. Process innovation$\left(P_I\right)$

Businesses can adjust to changes and uncertainty in the process parameters by introducing Fuzzy control into their process innovation control systems $P_I$. This versatility enables robust and optimized automation of the operational and production processes related to EI. $P_I$ involves adopting brand-new or vastly better production or distribution techniques and material modification methods, tools, and applications using the Eq (2).

$$P_I=\frac{P_{I}usageinalistofenterprises}{M_{total}}*100$$ (2)

Here, $P_I$ refers to Process Innovation either in a new way or improvised manner, and $M_{total}$ refers to the total count of the manufacturing enterprises. This $P_I$ involves production, logistics, distribution of goods, and process improvement for businesses, including Research and Development (R&D) activities, professional involvements, sales team, logistics suppliers, and communication services.Enterprise R&D investment mainly includes personnel investment and fund investment. Enterprise R&D investment and innovation performance should be considered the simplest input-output relationship. Here, the input innovative indicators may be R&D and marketing expenses, and the output may produce no. of new products with the newly added customers. This new addition is frequently updated in the Q-table. The process innovation control system may continually automate the operations and production processes and integrate Q-learning, resulting in more reliable and effective EI in all manufacturing firms.

3.1.2. Product innovation (${pdt}_I$)

Each enterprise has an innovation developer or driver for the products they innovated ${pdt}_I$ refers to the business that created and applied innovation for a business to create new or improved goods, which includes the tangible ones with design and services are intangible ones that may change according to the user time, concentration, communication, knowledge-oriented product attributes, and availability of the features of the service can be determined from these user based inputs; this can be done either inside or externally. Businesses may uncover the distinctive marketing elements that engage with clients and distinguish their goods in the marketplace by utilizing FL to assess and rank numerous product features and traits. The innovative product may be new to the market, new to the business, and partially modified.Thus, implementing a good or service that is novel or vastly improved in terms of its features or intended applications. Eq (3) comprises better technical requirements, parts, materials, software integration, ease of use, and other functionalities.

$${pdt}_I=\frac{Listofenterprisesusing{pdt}_I}{M_{total}}*100$$ (3)

The implementation of both $P_I$ and ${pdt}_I$ strategies in an enterprise, the proposed algorithm can improve resource efficiency ⍵ by using less material for each unit of production and lowering energy consumption $L_{ec}$ refers to the terms related to the production.Resource efficiency ⍵ is achieved through optimizing the production process and resulting in an efficient use of raw materials, thereby, reducing waste and improving overall resource efficiency. The decision-making process is guided by the Q-TSFC algorithm includes considerations for smart resource allocation using selective production innovation methods that maximize the output using minimal resources.The Q-learning aspect of the algorithm might identify and prioritize actions that lead to $L_{ec}$ that could involve optimizing the sequence of operations in the production workflow.

3.1.3. Innovation in organizational change$\left(O_I\right)$

Eq (4) involves following a new business follow-up, external relationship communication with suppliers like database and repairing hardware and software, business strategies like audits, finance, human resource management, and arranging the work responsibilities to employees and decision making. The Fuzzy control technique can consider these decision-making variables, each with a different weight, to reach the best decisions. Demand in the market, technological viability, and expected profit are considered. The most promising idea is then selected based on Fuzzy inference.

$$O_I=\frac{No.ofenterprisesusing{O}_I}{M_{total}}*100$$ (4)

3.1.4. Marketing innovation$\left({\Psi }_I\right)$

A Fuzzy Inference System (FIS)can predict future market circumstances more accurately by analyzing the previous data and market patterns. Informed strategic decisions about developing the product innovation, entry into the market, and position as competitors may be made by the businesses using this knowledge acquired by the training algorithms in an AI.

$$\left({\Psi }_I\right)=\frac{{\Psi }_{I}implementationinno.ofenterprises}{M_{total}}*100$$ (5)

Eq (5) describes the term ${\Psi }_I$ involves a new way of marketing or advertising innovation that involves the designing of an aesthetic feel of the product, promotional values of a product, telemarketing, pricing, Customer Relationship Management (CRM), and pre-and post-sales procedures. To maintain this, CRM enterprises perform customizing innovative options for every customer's unique requirement, increasing their satisfaction and engagement.

3.1.5. Cost savings from EI

The proposed technique can examine the enormous amounts of information and find patterns that employees might not notice. Innovation is related to cost savings related to the environmental advantages that can aid in reducing waste, lower energy consumption, minimizing delays, and optimizing processes, which will boost efficiency and result in cost savings. The Fuzzy linguistic variables for the benefits of cost savings due to EI activity show the {<5%-inadequate, 5–9.9%-less, 10-25%-average, 25-35%-sufficient, 35–50%profitable}

3.1.6. Classification of partnerships in the innovation collaboration

Co-operation is the engaged participation of other organizations in collaborative, innovative efforts. These could be other businesses or non-profit organizations, and the venture need not immediately provide the partners with financial gain. Co-operation does not occur when work is purely contracted out without active participation. Enterprises can obtain the knowledge and technologies through innovation co-operation that they couldn't otherwise use. Additionally, there is a lot of possibility for efficiencies in collaboration as partners benefit from one another. An enterprise or organization that collaborated with a different business on innovative projects is referred to as an innovation co-operation partner ${\zeta }_{\mathbb{P}}$. Partners are classified from the parental, material suppliers, private and public sector, competitors, academicians, non-profiters, and households.

$${\zeta }_{\mathbb{P}}=\frac{\%ofmanufacturingcompaniesengagedin\mathbb{P}}{M_{total}\left(I\right)}*100$$ (6)

Eq (6) defines the $\mathbb{P}$ that represents the partnership in co-operation with other entities. $M_{total}\left(I\right)$ refers to the total count of the manufacturing enterprises involved in the Innovation(I). Activities related to EI expenses involve all research & development and business endeavors conducted by a company to innovate and are categorized as innovation activities. It includes initiatives for new products, services, or business models that focus on licensing, training, machines, and tools. Employment-related expenses(labor) and investments for projects that has helped to develop and market fresh or better products, services, or company procedures are included in the innovation expenditures.

3.2. Implementation of Q-learning algorithm for predicting the innovation performance

The intelligent Q-learning approach that has been suggested indicates the efficiency or profitability of the chosen strategy in promoting innovation in businesses. Any relevant innovation indicator, such as increases in sales, satisfied customers, cost savings, market share, etc., could be used to evaluate this. The detailed procedure of Q-learning for identifying the innovation performance is given below.

Step1

Define the State Space ($s_t)$: The initial process describes the EI environmental features like the good's characteristics, process design, internal capabilities, marketing strategies, and attributes in the innovation environment.

Step 2

Define the Action Space $\left(a_c\right)$: Action results will correspond to the selection of particular characteristics and attributes to change the marketing campaigns, forming partnerships, or incorporate into the good or service related to the $P_I$ to improve the enterprise's performance and to achieve the cost savings.

Step 3

Setting up of Rewards ($\delta$): Based on the results of each action, provide the agent of the enterprises, namely the innovative driver's prizes. The cost savings and improved overall performance that the company has seen as a consequence of its activities should be reflected in rewards. Unfavorable rewards can deter behavior resulting in losses or inefficiencies, while positive awards should be offered for cost savings and performance improvements. The outcome of rewards based on the chosen features and traits reflects the success of the market trend and the satisfaction of the customers.

Step 4

Calculation of values in Q-table: A matrix called the Q-table links state and behaviors to the corresponding Q-values. The projected cumulative benefits that the company can obtain by performing a given action in a specific scenario are represented by Q-values. Random values are used to initialize the Q-table. The predicted cumulative rewards like EI performance, which the enterprise can achieve by doing a certain action in a specific state, are represented by the Q-values stored in the Q-table for each state and action $\left(s_t,a_c\right)$ combination in Eq (7).

$$Q\left(s_t,a_c\right)=\left(1-\lambda \right)*Q\left(s_t,a_c\right)+\lambda *\left(\delta +\gamma *\max \left(Q\left(s_t^{\prime },{a_c}^{\prime }\right)\right)\right)$$ (7)

Where $\mathrm{\lambda }$ represents the learning rate, $\mathrm{\gamma }$ defines the discount factor, the next state and action represent the ${\mathrm{s}}_{\mathrm{t}}^{\prime },{{\mathrm{a}}_{\mathrm{c}}}^{\prime }$.The term $\max \left(Q\left(s_t^{\prime },{a_c}^{\prime }\right)\right)$ refers to the target policy called the greedy approach.Update the Q-values throughout the learning process, and the agent can use this calculation to update its knowledge of the anticipated benefits of every state-action pair $\left(s_t,a_c\right)$ based on its experiences.

Step 5

Exploration-Exploitation Trade-Off using the Greedy Approach: Based on the Q-values discovered thus far, the ε-greedy strategy aids the innovation developer in an enterprise in finding a balance between investigating novel possibilities and taking advantage of the most well-known actions that give success in the market place which gives profit to an organization and satisfies the customer needs.

Step 6

Training Result of Rewards ($\delta$): The chances of exploring a random action rather than selecting the one with the highest Q-value are determined by the rate of exploration. Higher levels of ε results encourage more investigation, enabling the enterprise to seek novel, creative strategies more frequently. The emphasis moves to utilize the more profitable actions based on the learning Q-values as they decline with the period or as the innovation developer, called an agent in an enterprise, gains more knowledge. The agent engages in repeated interactions with its surroundings, acting and being rewarded. The agent learns to make decisions that result in greater $\delta$ and cost savings through the multiple training instances, which update the Q-values.

Step 7

Learned Q-values through the optimal training policy: The Q-learning process convergence occurs to an efficient, innovative strategy after a significant amount of training, at the point the agent or innovative developer determines the best course of action to adopt in every state for maximizing rewards and accomplish the cost-saving goals.

The proposed intelligent Q-learning approach demonstrates the effectiveness or profitability of the selected strategy in fostering innovation in enterprises. It could be assessed using any appropriate innovation indicators, such as revenue growth, customer happiness, cost reductions, share of the market, etc. When Fuzzy control, as well as reinforcement learning, are combined, the enterprise can make updated and flexible decisions regarding all the features related to enterprise innovation, resulting in the development of new or significantly improved products with improved technical specifications, user interfaces, designing, implementation of the software, and other features.

3.3. Takagi sugeno fuzzy control application in enterprise innovation

The output of the Q-learning process provides a quantitative assessment of the desirability of various innovative actions in different circumstances once the Q-values have been learned. These Q-values indicate the activities are more likely to result in higher rewards in the particular market states, which encapsulate the knowledge acquired through the learning process of all innovation strategies through an optimal policy. The Q-values produced from Q-learning are one of the inputs used by the Fuzzy control mechanism. The Fuzzy control system subsequently guides the decision-making process, incorporating this quantitative data with linguistic guidelines and the fuzzy membership functions. The Fuzzy control framework evaluates the Q-values and other variables to generate judgments and suitable activities based on the Fuzzy basis rules.The Q-values improve with time and represent the company's changing understanding of the innovation environment.The organization's innovation strategy can be adjusted to change the market dynamics and the environmental conditions thanks to the learned Q-values. The Q-values direct the organization towards more fruitful and pertinent innovation paths as the environment changes.The company may still consider developing new products to exploit the market's high demand and enough supply. Fig. 3 depicts the implementation procedure of the TSFC algorithm for making decisions on EI.

Fig. 3.

The implementation process of TSFC in enterprise innovation decision-making.

The organization can predict the innovation results more precisely by utilizing Q-learning's predictive skills and the Fuzzy control's ambiguity management. The obstacles that hinder the EI include resource limits, external and market. The lack of knowledge among the employees and the lack of capital are resource-constrained. The external factors are logistics, and the market factors include suppliers, foreign customers, personalization of the customer demands, and uncertainty. Fuzzy linguistic terms are defined as {very hard, hard, somewhat hard, not at all hard, not applicable} to handle this uncertainty issues of EI obstacles using the proposed approach. The formulas used in the research content are checked thoroughly for explanation, and the symbol description is given in Table 2.

Table 2.

List of symbols.

Symbols	Explanation
$I_R$	Innovation Ratio
$P_I$	Process Innovation
$M_{total}$	total count of the manufacturing enterprises
${pdt}_I$	Product Innovation
⍵	resource efficiency
$L_{ec}$	lower energy consumption
$O_I$	Innovation in organizational change
${\Psi }_I$	Marketing Innovation
${\zeta }_{\mathbb{P}}$	co-operation partner
$s_t$	State space
$a_c$	Action space
$\lambda$	Learning rate
$\delta$	Rewards
$\gamma$	Discount factor
$\left(s_t,a_c\right)$	state-action pair
$s_t^{\prime },{a_c}^{\prime }$	Represents next state and action
$Q\left(s_t^{\prime },{a_c}^{\prime }\right)$	Target Policy
ε	Greedy strategy
$A$	universe of the disclosure
$x$	input variables
${\mu }_{{pdt}_I}\left(A\right)$	membership function for the product innovation
$R^2$	coefficient of determination

Fuzzy language variables and reasoning provide the foundation of Fuzzy control. According to the study of the fundamental concepts, Fuzzy control uses its system control on a Fuzzy representation of the controlled object. The linguistic ones are transformed from actual variables throughout the fuzzification process. The enterprise innovators choose the factors that require to be considered in the knowledge base for an effective innovation outcome like productivity improvement. The fundamental advantage of FLCs is that phrases rather than equations may be used to operate a system, also known as a process. Linguistic rules can be used to explain a control strategy in a manner more akin to human language. Particularly, when the processes are too complicated for study by standard approaches or whether the information accessible sources are interpreted qualitatively, precisely, or uncertainly, the FLC methodology proves to be highly helpful. Crisp sets are fuzzy sets when the membership function has been turned into its characteristic variables of enterprise unit with a range from [0,1]. For an element x, the universe of the disclosure may be A, which consists of input variables x ϵ A, where the A may involve the innovation criteria like product, process, organization, and market based on the membership values of 0,1. As mentioned above, 0 represents no association, and one indicates highly associated. The term membership function refers to the mathematical function in a triangular representation representing each innovation driver, the expert's decisions, and stakeholders' feedback through the questionnaire collected. For each innovation criteria, the linguistic terms defined are x⟶ {very low:0.0–0.3, low:0.3–0.5, moderate:0.5–0.7, high:0.7–1.0} for product innovation the membership function can be represented as ${\mu }_{{pdt}_I}\left(A\right)$.If the crisp value represents the ${pdt}_I$ with numerical values are x⟶ {very low:0.2, low:0.4, moderate:0.6, high:0.9}.The corresponding membership function is depicted in Fig. 4.

Fig. 4.

Membership degree of ${\mathit{p}\mathit{d}\mathit{t}}_{\mathit{I}}$ Levels.

FL aids in decision-making by helping to identify the uncertainties in a situation and adapting to changing circumstances in the enterprise environment. The control approach is determined by a set of Fuzzy rules that follow an IF-THEN syntax in the rule base. By employing linguistic words, the Fuzzy control algorithm aids in interpreting and applying the learned production innovation policies in a more understandable and human-readable manner. The implementation of the innovation strategies can be analyzed by the expert system with the linguistic terms whether the uncertainties of innovative ideas are suitable to implement in the market with the cautions mentioned in the intelligent Q-learning application is given by {A: Highly preferable B: excellent, C: good, D: reasonable, E: Warning, F: too inconsistent}. The sample control decision outcome for EI is written in the form of a Fuzzy rule.

• 1)If the $P_I$ is B AND the ${\Psi }_I$ is D, THEN the control decision is C
• 2)If ${pdt}_I$ is F AND the $O_I$ is F, THEN the control decision is E
• 3)If resource allocation is C AND customer satisfaction is A, THEN the fuzzy control decision is B.

Because of its versatility, TSFC may modify its decision-making processes in response to immediate inputs from Q-learning and changes in the innovation atmosphere. This adaptability ensures that decisions are made in light of the context and response to shift market trends. Because the linguistic rules of TSFC are transparent, businesses may understand the assess innovation performance. It enables the targeted improvements and the optimization of the innovation efforts by evaluating the contribution of the numerous aspects to the overall performance. The defuzzification technique can be used to get a clear judgment or value centered on the imprecise outcomes of the inference Fuzzy system. The defuzzification procedure can transform the Fuzzy output indicating the quality of innovativeness into a definite numerical number that represents the level of creativity necessary for a given product. The TSFC algorithm is implemented to make judgments relating to product innovation.

4. Results and discussion

4.1. Data source-overview

A cross-economy assessment of the Canadian business enterprises and commercial organizations that are not-for-profit is termed as the assessment of innovation and business strategies taken from the data sources of [31,32] with small (20–99), medium (100–249), and the large enterprises (250). The following 14 sectors, as defined by the North American Industry Classification System, are the only ones whose businesses can participate in the Survey of Innovation and Business Strategy related to EI. Only enterprises possessing a minimum of twenty workers and profits of a minimum of $250,000 were considered for choosing the sample to lessen the response weight on smaller enterprises, considering employee size, location, and industry. The list of the industrial sectors with code and relevant variables for innovation activities is discussed in Table 3.

Table 3.

List of industrial sectors with code for relevant EI activities.

Industrial sectors	Enterprise codes
Agriculture, forest, hunting, fishing	11
Mining, quarry, oil and gas extract	21
Utilities	22
Construction	23
Manufacturing	31–33
Wholesale trading	41
Retailing	44–45
Transportation &Warehouse	48–49
IT & cultural sectors	51
Finance(bank)	52
Real-estate	53
Academic, professional services	54
Enterprise management	55
Administration	56

4.2. Comparative study

The Comparative study validates the performance of the Q-TSFC method in contrast to existing techniques, including AI-ENFL [21], B-AI [23], and QXGB [24] for comparison of EI in various industrial sectors with its performance metrics.

4.2.1. Customer satisfaction ratio with an EI strategies

Fig. 5 shows the customer satisfaction ratio based on the enterprise's innovative strategies updated by using historical data and consumer feedback to discover the efficient innovation actions by enterprises that align with the customer needs; Q-learning supports data-driven decisions. With the use of the TSFC, innovation efforts like designing a product customized depending on the linguistic conventions and customer preferences allow for the customization of goods and services to suit specific requirements. The integrated strategy encourages adaptability and continual improvement with frequent updates to respond quickly to shifting market conditions and consumer expectations. Optimized decision-making from Fuzzy control leads to fewer errors and better client experiences. Utilizing cutting-edge products and customer-focused tactics results in an outstanding position in the marketplace, boosting client confidence and happiness. Using client input to fix issues and generate continual changes promotes customer involvement over time and contributes to continuing the business success.

Fig. 5.

Customer satisfaction ratio on EI.

4.2.2. Calculation of enterprise performance efficiency (%)

Fig. 6 illustrates the performance efficiency analysis of an enterprise by implementing AI-based innovative strategies, and enterprises can access cutting-edge technologies, enhance innovation capabilities, and improve overall efficiency.Enterprises can combine their capital, knowledge, and expertise through partnerships. Working with partners gives business organizations access to specialized knowledge and expertise in multiple industries. The internal competencies of the enterprise may be supplemented by this expertise, resulting in more innovative and effective decision-making with the Fuzzy control approach.As a result, innovation procedures are streamlined through process, product, organizational, marketing, cost savings, and partnershipbased on the strategies through which response times are sped up, which boosts productivity and lowers operating expenses. The response time may involve an organization's time to respond to an incident (market change), resulting in operational efficiency, customer satisfaction, and overall performance.

Fig. 6.

Enterprise performance efficiency analysis on various strategies.

4.2.3. Analysis of cost savings on different algorithms (%)

Q-learning and Fuzzy control-driven innovation strategies can increase efficiency in developing goods and services through other businesses. Increased productivity entails producing more with the minimal number of resources utilized for innovations, which saves money. These cost savings are represented as linguistic terms through which decision towards the enterprise profitability analysis is performed like {<5%-inadequate, 5–9.9%-less, 10-25%-average, 26-35%-sufficient, 35–50% profitable}.

Fig. 7 depicts the cost savings scheme based on the various innovation strategies in enterprises compared to existing algorithms. The proposed Q-TSFC approach produces a profitable amount of savings in this innovation process, followed by AI-ENFL [21] as a sufficient amount of savings, the B-AI [23] gives the acceptable average expenses, the QXGM [24] approach gives the least performance among all other algorithms. The organization can achieve its innovation goals at a lesser cost by concentrating on inventions that have greater rewards and less associated expenses. Lower energy consumption, decreased waste, and less resource utilization-focused innovations can result in long-term economic savings and environmental advantages.The comparison of AI-ENFL, B-AI, and QXGB algorithms sheds light on the cost-effectiveness of different approaches, with Q-TSFC demonstrating superior performance.

Fig. 7.

Cost-saving analysis of each algorithm on various innovations.

4.2.4. Performance comparison of R2

Employing the Q-TSFC approach, R² value is critical in directing data-driven decision-making in each facet of the innovation (product, process, market, organizational, and cost savings). The most effective innovation variables are identified, resource allocation is optimized, and innovation tactics are continuously improved. Enterprises can improve their capacity for Innovation, generate effective results, and keep a competitive advantage in the market by utilizing $R^2$ as a component of the evaluation process.

$$R^2=1-\left(\sum {\left(predicted-mean\right)}^2/{\left(actual-mean\right)}^2\right)$$ (8)

The above Eq (8) defines the dependent and independent variables that are calculated for all data points related to the innovation variables; the dependent variables include the response outcome of the innovation success rate, customer satisfaction ratio, cost savings, and several other performance indicators, which vary based on the independent variables called predictors that are observed by Q-learning approach and controlled by Fuzzy principles linguistic terms with the findings of uncertain market trends, process changes, organizational competencies.The correlation between dependent and independent variables showcases the capability of the Q-TSFC approach to make meaningful predictions about the performance state of EI.

In Fig. 8, a high R² value with the regression analysis means that the EI framework is good at predicting which product changes will be well-received by consumers, increase process efficiency, meet market demands, improve productivity, customer and employee satisfaction, highest returns by the cost savings, and lead to the desired results. Suppose the dependent and independent variables are strongly correlated, as shown by a high R2 value. In that case, making predictions and reaching meaningful conclusions about the EI's performance state may be possible. The strong correlation between dependent and independent variables indicates the potential of the Q-TSFC approach to make meaningful predictions about the performance state of enterprise innovation.

Fig. 8.

Analysis of R2 for different algorithms.

The comparative study showcases the strengths of the Q-TSFC method in enterprise innovation, particularly in terms of customer satisfaction, performance efficiency, cost savings, and predictive capabilities. The adaptability and integration of Q-learning with fuzzy control contribute to more flexible and effective decision-making in response to dynamic market conditions. The emphasis on partnerships and collaboration further underlines the potential for enhancing internal competencies through external expertise.The comparative result findings suggest that the Q-TSFC approach offers a promising solution for enterprises seeking to optimize their innovation strategies.This comprehensive comparative discussion provides a nuanced understanding of the Q-TSFC method's advantages over existing techniques like AI-ENFL,B-AI, and QXGB, offering valuable insights for researchers in the field of enterprise innovation.

5. Conclusions

Applying AI and Fuzzy control techniques in EI can completely transform the way firms operate and succeed in a relentless market. The proposed Q-TSFC algorithm is truly beneficial where combines adaptive learning with decision-making leading to easy comprehension of human beings. Q-learning encourages continuous improvement with the updated table. It is helpful when new market predictions are received, and customer demands are modified over the years, insights are gathered along the innovation journey, resulting in improved methods for innovation and overall performance. Businesses can improve their decision-making processes by utilizing Q-learning through continuous acquiring knowledge from rewarding their experiences and relationships with the innovation environment. TSFC is incorporated allowing for flexible language decision rules and comprehensible decisions that align with human competence.The main goal of Q-TSFC approach is to deal with language imprecision and uncertainty in market trends, allowing businesses to make intelligent choices despite the lack of accurate numerical data. The technique improves the distribution of innovation related to partnership, productivity, customer happiness, and innovative performance efficiency through the adaptive acquisition of knowledge and Fuzzy logic. These enhancements result in cost reductions and market advantage.In the proposed Q-TSFC algorithm, the obtained results are 96.5% customer satisfaction ratio, 96% enterprise performance efficiency, cost savings of about 48% profitable value, and the coefficient of determination R2 is 0.83, respectively.The limitation of this work is accurate and exhaustive data can be difficult to collect in real-world situations, which could result in inaccurate learning of Q-values and Fuzzy constraint interpretations. The future direction may involve businesses accessing real-time data about customer behaviour, market trends, and operational performance by integrating big data streams. More reliable and contextually aware decision-making can result from incorporating these data sources into the Q-TSFC algorithm.

Data availability statement

The data used to support the findings of this study are all in the manuscript.

Funding statement

The key project of philosophy and social science planning of zhejiang province (23NDJC055Z).

CRediT authorship contribution statement

Yanhuai Jia: Conceptualization, Formal analysis, Writing – original draft. Zheng Wang: Formal analysis, Methodology, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no competing interests.