Abstract
Background
This study aimed to develop and validate a risk prediction model for smartphone addiction among college students using an extreme gradient boosting (XGBoost) algorithm, and to identify key factors associated with this behavioral pattern.
Methods
A cross-sectional survey was conducted among 2,761 college students. The XGBoost machine learning algorithm was applied to analyze the dataset, enabling the identification of variables associated with smartphone addiction and the ranking of their relative contributions based on feature importance scores.
Results
The prevalence of smartphone addiction in this sample was approximately 22.24%. According to the XGBoost model, predictors ranked by descending feature importance scores were as follows: Loneliness (0.437), Monthly household income (0.067), Age (0.056), Place of residence (0.056).
Conclusions
Smartphone addiction among college students represents a notable public health concern requiring sustained attention. University educators and administrators are encouraged to prioritize factors showing robust statistical associations with smartphone addiction—including loneliness, monthly household income, age and place of residence—implement routine screening for problematic smartphone use, and develop personalized intervention strategies aligned with individual risk profiles.
Keywords: college students, machine learning, predictive modeling, smartphone addiction, XGBoost
Introduction
Smartphone addiction, also referred to as smartphone dependence, problematic smartphone use, excessive smartphone use, or nomophobia, is a behavioral disorder characterized by compulsive smartphone engagement that significantly impairs daily functioning, interpersonal relationships, occupational performance, and overall well-being (1, 2). Although not currently classified under the “Substance-Related and Addictive Disorders” category in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5), smartphone addiction is widely recognized as a form of behavioral addiction (3). It exhibits core diagnostic features consistent with established addiction models, including salience, tolerance, withdrawal symptoms, interpersonal conflict, and relapse (4, 5). Notably, accumulating evidence indicates substantial overlap between behavioral and substance addictions across clinical presentations, cognitive processes, neurobiological mechanisms, and neuroimaging profiles (6).
Given its high prevalence, rapid escalation, and substantial adverse consequences, smartphone addiction has emerged as a global public health concern (7). This issue is particularly pronounced among university students in China. A national survey reported that approximately 25% of Chinese college students exhibit problematic smartphone use (8). Compared with other populations, university students typically possess greater discretionary time (9), yet often demonstrate psychological immaturity and limited self-regulatory capacity—factors that collectively heighten vulnerability to smartphone addiction. Prolonged excessive smartphone use has been associated with detrimental effects on both physical and mental health, including poor sleep quality and elevated depressive symptoms (10, 11). In light of the widespread smartphone engagement and its documented harms in this population, elucidating the underlying triggers and mechanistic pathways is critically important for developing targeted prevention and intervention strategies. Although existing studies have investigated the factors influencing smartphone addiction among college students, the majority primarily rely on conventional regression models for analysis (12–15). These methods typically operate under linear or generalized linear assumptions, making it difficult to effectively capture complex non-linear relationships and higher-order interactions among variables. Furthermore, they are sensitive to multicollinearity, resulting in limited predictive accuracy and model generalizability. XGBoost is an advanced ensemble machine learning algorithm that constructs base learners by optimizing a regularized objective function. To enhance computational efficiency and generalization performance while effectively mitigating overfitting, the algorithm incorporates techniques such as pre-sorting and weighted quantile sketching (16). SHAP (Shapley Additive exPlanations) values, grounded in cooperative game theory, quantify the magnitude and direction of each feature’s contribution to model predictions, thereby providing an intuitive and interpretable assessment of feature importance (17). Within this methodological framework, the present study employs an XGBoost model integrated with SHAP-based interpretation to systematically investigate the determinants and interaction patterns underlying smartphone addiction among college students. This approach aims to deliver actionable insights for educators and student support professionals, facilitate the early identification of high-risk individuals, and inform the development of targeted prevention and intervention strategies to mitigate the prevalence and adverse impacts of smartphone addiction in this population.
Methods
Between March and April 2026, college students were recruited from five universities across Shandong, Hunan, and Hebei provinces using a convenience sampling approach. Data were collected via anonymous self-administered questionnaires distributed by class advisors. Recruitment and data collection continued until no new submissions were received for seven consecutive days.
Inclusion criteria: (1) full-time enrollment at a participating university; (2) sufficient proficiency in Mandarin Chinese to independently comprehend and complete the survey; and (3) willingness to provide informed consent.
Exclusion criteria: (1) a documented history of psychiatric disorders; and (2) being on medical leave or academic suspension during the data collection period.
General information and questionnaire data
Demographic characteristics
Baseline sociodemographic characteristics were recorded across 11 items, including gender, age, et al. educational level, mother’s education level and father’s education level et al.
Smartphone Addiction Scale-Short Version
Smartphone addiction was assessed using the SAS-SV, originally developed by Kwon (18) et al. and subsequently translated and culturally adapted into Chinese by Xiang (19) et al. The scale comprises 10 items rated on a 6-point Likert scale, ranging from 1 (strongly disagree) to 6 (strongly agree). Total scores range from 10 to 60, with higher scores indicating more severe smartphone addiction. In the present study, a total score >33 was used as the cutoff to identify smartphone addiction. The SAS-SV demonstrated excellent internal consistency reliability in this sample, with a Cronbach’s α coefficient of 0.878.
The University of California at Los Angeles Loneliness Scale (UCLA-20)
UCLA-20, developed by Russell et al. (20), was used to assess loneliness. The scale comprises 20 items, 9 of which are reverse-scored. Responses are rated on a 4-point Likert scale ranging from 1 (“never”) to 4 (“always”), yielding a total score ranging from 20 to 80. Higher scores indicate greater levels of loneliness. Consistent with previous studies (21), a cutoff score of ≥44 was adopted to classify participants as experiencing loneliness. In the present study, the internal consistency of the UCLA-20 was excellent, with a Cronbach’s α coefficient of 0.870.
Data collection
Data collection was conducted using an electronic questionnaire. The survey link or QR code was distributed to college students via class counselors. Following a clear explanation of the study objectives, anonymity and confidentiality measures, and the voluntary nature of participation, participants provided informed consent and completed the survey independently. All responses were automatically submitted in real time to the online database. Should any participant encounter difficulties with question comprehension or platform navigation, research staff provided strictly neutral, non-directive technical assistance to ensure data accuracy without introducing response bias.
Sample size
The sample size was determined based on the events-per-variable (EPV) principle, which recommends a minimum of 10 outcome events per predictor variable to ensure model stability (22). Accounting for an anticipated attrition rate of 10%–20% and 11 candidate variables planned for multivariable modeling, the minimum required sample size was calculated to be 122 participants. A total of 2,785 questionnaires were distributed, and 2,761 valid responses were obtained, yielding a valid response rate of 99.14%.
Statistical analysis
This study was conducted in a Python 3.9 environment, utilizing the Pandas, NumPy, and XGBoost libraries for predictive modeling and statistical computation. After rigorous data cleaning, the analytical dataset comprised 2,761 complete observations across 11 predictor variables (including both continuous and categorical features) with no missing values.To reduce model complexity, enhance interpretability, and mitigate overfitting risks, we first employed LASSO (Least Absolute Shrinkage and Selection Operator) regression for key variable selection, leveraging its L1-regularization property to shrink irrelevant coefficients toward zero. The dataset was randomly split into a training set (80%) and an independent test set (20%). All modeling procedures—including feature selection, hyperparameter optimization, and model training—were performed exclusively within the training set using 5-fold cross-validation to ensure robustness; the test set was held out completely and used solely for unbiased evaluation of the final model performance. Based on the selected features, we systematically developed and compared seven machine learning models: Logistic Regression, Elastic Network, K-Nearest Neighbors (KNN), Decision Tree, XGBoost, Support Vector Machine (SVM), and Random Forest (see Supplementary Materials). The XGBoost model, which demonstrated the best overall performance in terms of accuracy, AUC, and F1-score, was ultimately selected to predict and stratify smartphone addiction severity. XGBoost (Extreme Gradient Boosting) minimizes prediction errors by iteratively constructing decision trees in a stage-wise additive manner, where each new tree corrects the residual errors of its predecessors. Its key hyperparameters—including maximum tree depth (max_depth), number of base learners (n_estimators, representing the total number of sequential decision trees), and learning rate (learning_rate, controlling the contribution weight of each tree)—were comprehensively optimized via grid search combined with 5-fold cross-validation within the training set to ensure robust generalization. To strictly prevent information leakage—a critical concern in machine learning pipelines—the feature selection process was fully nested within the cross-validation loop: in each fold, the LASSO model was independently fitted and variables were selected using only the current training subset without any exposure to validation or test data; the selected variables were then consistently applied to both the training and validation subsets of that fold, while the test fold data were completely excluded from all preceding modeling steps, including feature screening and hyperparameter tuning. Following systematic optimization, max_depth = 2 and n_estimators = 26 (i.e., 26 base learners) were identified as the optimal parameters to balance model performance, computational efficiency, and interpretability. The final model was selected based on the highest mean accuracy across all cross-validation folds, with additional consideration of metric stability (low standard deviation across folds).To enhance model interpretability and bridge the gap between black-box predictions and clinical decision-making, SHAP (Shapley Additive exPlanations) analysis, grounded in cooperative game theory, was employed to quantify the magnitude and direction of each feature’s contribution to individual predictions. The cross-validation-guided feature selection process effectively stabilized feature importance rankings across resampling iterations, ensuring the selection of an optimal and reproducible feature subset, thereby simultaneously improving predictive accuracy, model transparency, and clinical interpretability for practical implementation.
Results
Demographic characteristics of participants
A total of 2,761 college students were enrolled in this study, comprising 863 males and 1,898 females. Among them, 614 participants (22.24%) were identified as having smartphone addiction. Detailed demographic characteristics are presented in Table 1.
Table 1.
Demographic characteristics of participants.
| Variable | Total | No | Yes |
|---|---|---|---|
| Gender | |||
| Male | 863 | 660 | 203 |
| Female | 1898 | 1487 | 411 |
| Age(Years) | |||
| <19 | 1481 | 1139 | 342 |
| 20-22 | 1204 | 944 | 260 |
| >22 | 76 | 64 | 12 |
| Educational Level | |||
| Associate Degree Student | 1291 | 969 | 322 |
| Undergraduate Student | 1470 | 1178 | 292 |
| Mother’s Education Level | |||
| Junior High School or Below | 1625 | 1259 | 366 |
| Senior High School or Vocational Secondary | 849 | 661 | 188 |
| College or Above | 287 | 227 | 60 |
| Father’s Education Level | |||
| Junior High School or Below | 1371 | 1064 | 307 |
| Senior High School or Vocational Secondary | 1028 | 804 | 224 |
| College or Above | 362 | 279 | 83 |
| Place of residence | |||
| Rural | 1441 | 1127 | 314 |
| Urban | 1320 | 1020 | 300 |
| Monthly Income(¥) | |||
| <3000 | 746 | 561 | 185 |
| 3001-6000 | 1239 | 985 | 254 |
| 6001-9000 | 488 | 386 | 102 |
| ≥9001 | 288 | 215 | 73 |
| Live on campus | |||
| No | 50 | 34 | 16 |
| Yes | 2711 | 2113 | 598 |
| Smoking | |||
| No | 2616 | 2052 | 564 |
| Yes | 145 | 95 | 50 |
| Monthly living expenses(¥) | |||
| <1000 | 767 | 597 | 170 |
| 1001-2000 | 1895 | 1471 | 424 |
| >2000 | 99 | 79 | 20 |
| Loneliness | |||
| No | 1715 | 1505 | 210 |
| Yes | 1046 | 642 | 404 |
Risk factors for smartphone addiction among college students
The LASSO regression analysis is illustrated in Figures 1 and 2. The optimal regularization parameter (λ) was determined to be 0.013 through cross-validation. Based on this parameter, six predictor variables with non-zero regression coefficients were selected for the final model: Gender, Age, Place of residence, Monthly household income, Smoking, and Loneliness (Table 2).
Figure 1.
Lasso regression coefficient paths for predictors of smartphone addiction.
Figure 2.
Mean Squared Error (MSE) path with cross-validation for selecting the optimal regularization parameter.
Table 2.
Results of LASSO regression for predictors associated with smartphone addiction in college students.
| Variables | Coefficient | Lambda. min |
|---|---|---|
| Gender | -0.013 | 0.013 |
| Age | -0.006 | |
| Place of residence | 0.008 | |
| Monthly Household Income | -0.005 | |
| Smoking | 0.077 | |
| Loneliness | 0.256 |
XGBoost model results
Optimal feature subset
Features selected by LASSO regression (non-zero coefficients) served as input variables for the XGBoost classifier. We employed 5-fold cross-validation to systematically evaluate model performance across different feature combinations. Classification accuracy (range: 0–1) was used as the evaluation metric, with a 4-feature subset achieving optimal performance (Accuracy=0.706, Precision=0.638, Specificity=0.873, F1-Score=0.494, 95%CI=0.604-0.643), see Figure 3.
Figure 3.
Line graph of the optimal feature subset screening.
Feature importance ranking
Categorical variables were encoded as follows: Loneliness (0 = No, 1 = Yes); Monthly household income (0 = ≤3000, 1 = 3001–6000, 2 = 6001–9000, 3 = ≥9000); Age (0 = <19 years, 1 = 20–22 years, 2 = ≥22 years); and Place of residence (0 = Rural, 1 = Urban). The factors influencing smartphone addiction among college students, ranked by mean absolute SHAP values, are: Loneliness (0.437), Monthly household income (0.067), Age (0.056), Place of residence (0.056). The feature levels Loneliness (Yes), monthly household income (≥9000 RMB), age (≥22 years), and place of residence (Urban) all demonstrated strong positive SHAP values, suggesting that these factors substantially elevate the predicted risk of mobile phone addiction in the college student population (Figures 4, 5).
Figure 4.
Feature importance ranking.
Figure 5.
SHAP value (impact on model output). X-axis (SHAP value): Positive values (right side) → Increased predicted risk of mobile phone addiction. Negative values (left side). Decreased predicted risk of mobile phone addiction.Color bar:Blue = Low feature value (lower encoded value). Red = High feature value (higher encoded value).
Discussion
The prevalence of smartphone addiction among college students in this study was 22.24%, slightly lower than the 28.49% reported by Wang et al (23). This discrepancy may be attributable to differences in assessment instruments and regional characteristics. Nevertheless, the observed prevalence remains notably high. Existing literature indicates that smartphone addiction is associated with compromised sleep quality, including insomnia and disrupted sleep patterns. Additionally, prolonged screen time has been linked to elevated psychological stress and co-occurring negative affective states, such as anxiety and depressive symptoms, underscoring potential implications for mental well-being (23). To elucidate variables associated with smartphone addiction and inform targeted intervention strategies, we applied the XGBoost algorithm to identify and rank predictive features. Feature importance analysis revealed the following predictors, ordered by their relative contribution to the model: loneliness, monthly household income, age and place of residence.
Our findings indicate that loneliness is significantly associated with smartphone addiction among college students, a result consistent with previous cross-sectional studies (24, 25). Loneliness is conceptualized as a multifaceted emotional state that emerges when an individual’s needs for intimacy and social connection are inadequately met. According to the Evolutionary Theory of Loneliness (ETL) (26), humans have evolved an inherent drive to maintain social bonds. Perceived social disconnection triggers an aversive emotional experience, which theoretically functions as an adaptive signal motivating individuals to restore and sustain interpersonal ties. As a highly accessible communication medium, smartphones may offer a readily available channel for social engagement. Consequently, college students experiencing loneliness may turn to their devices to facilitate interpersonal connection. This pattern aligns with Media System Dependency Theory (27), which posits that individuals develop reliance on specific media channels when they perceive them as essential or optimal for achieving personal goals, such as social integration. Given the immediacy and convenience of smartphones in fulfilling social needs compared to alternative media, individuals reporting higher levels of loneliness may exhibit a greater tendency toward problematic smartphone use.
Additionally, our findings indicate that monthly household income is significantly associated with smartphone addiction among college students, a result consistent with prior research (28). Students from higher socioeconomic backgrounds typically gain earlier exposure to smart devices and have greater access to feature-rich, high-end smartphones (29). Against the backdrop of rapid advancements in mobile communication technology, high-end devices offer increasingly diverse functionalities and interactive capabilities, which may facilitate prolonged engagement and higher usage frequency. Theoretical perspectives on media and behavioral engagement suggest that enhanced device functionality and accessibility are linked to patterns of intensive use. Our findings further indicate that age is a significant predictor of smartphone addiction among college students. This association may be partially attributable to developmental differences in self-regulation and academic habits. Moreover, the widespread integration of online learning has introduced environmental vulnerabilities; the reduced external supervision inherent in digital classrooms may impair students’ ability to self-monitor device usage (30). College students from urban areas typically experience earlier smartphone adoption and higher device accessibility, which may foster greater daily reliance and more frequent usage patterns. Moreover, as smartphones have become a primary medium for peer socialization, their constant provision of novel information and seamless global connectivity may collectively contribute to an elevated risk of smartphone addiction among urban undergraduates (30).
Strengths and limitations
This study employed an XGBoost-based approach to predict the risk of smartphone addiction among college students using data from a multi-center survey. Nevertheless, several limitations warrant acknowledgment. First, the reliance on convenience sampling and self-reported measures may introduce selection and recall biases. Second, the cross-sectional design precludes causal inferences regarding the relationships among the examined variables. Third, the predictive model was constrained by a limited set of covariates, and potential class imbalance within the dataset may have affected performance metrics. Additionally, although cross-validation was implemented, the absence of an independent external validation cohort limits the generalizability of the findings. There is also a possibility of information leakage during the feature selection process (e.g., if screening was conducted prior to data splitting), which may have contributed to an overestimation of model performance. Finally, the study did not incorporate biological markers (e.g., genomic data) or certain behavioral covariates (e.g., physical activity), thereby limiting insights into underlying physiological mechanisms and potential confounding pathways. Future research should prioritize longitudinal designs, recruit larger and more diverse cohorts for external validation, and integrate multi-dimensional data (including behavioral and biological indicators) to elucidate causal mechanisms and enhance predictive robustness.
Conclusion
In this study, an XGBoost model was employed to identify factors associated with smartphone addiction among 2,761 college students. The key predictors identified included loneliness, monthly household income, age and place of residence. Our findings underscore the importance of early psychological screening and assessment for college students by educators and university administrators. Furthermore, targeted supportive interventions should be implemented to address modifiable risk factors, thereby mitigating the risk of smartphone addiction within this population.
Funding Statement
The author(s) declared that financial support was not received for this work and/or its publication.
Footnotes
Edited by: Yu Tian, Qingdao University of Science and Technology, China
Reviewed by: Zamira Shabani, Luigj Gurakuqi University of Shkodër, Albania
Evi Karali, University of West Attica, Greece
Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.
Ethics statement
The studies involving humans were approved by Research Ethics Committee of Yanshan University (2021Y01). The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.
Author contributions
YL: Writing – review & editing, Methodology, Writing – original draft, Software, Investigation. NZ: Methodology, Writing – review & editing, Investigation, Writing – original draft. YW: Supervision, Investigation, Writing – review & editing, Formal analysis, Writing – original draft.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpsyt.2026.1850706/full#supplementary-material
References
- 1. Busch PA, Mccarthy S. Antecedents and consequences of problematic smartphone use: A systematic literature review of an emerging research area. Comput Hum Behav. (2021) 114:106414. doi: 10.1016/j.chb.2020.106414 38826717 [DOI] [Google Scholar]
- 2. Nawaz S. Rethinking classifications and metrics for problematic smartphone use and dependence: Addressing the call for reassessment. Comput Hum Behav Rep. (2023) 12:100327. doi: 10.1016/j.chbr.2023.100327 38826717 [DOI] [Google Scholar]
- 3. First MB. Diagnostic and statistical manual of mental disorders, 5th edition, and clinical utility. J Nerv Ment Dis. (2013) 201:727–9. doi: 10.1097/NMD.0b013e3182a2168a [DOI] [PubMed] [Google Scholar]
- 4. Csibi S, Griffiths MD, Demetrovics Z, Szabo A. Analysis of problematic smartphone use across different age groups within the 'components model of addiction'. Int J Ment Health Addict. (2021) 19:616–31. doi: 10.1007/s11469-019-00095-0 [DOI] [Google Scholar]
- 5. Sassover E, Weinstein A. Should compulsive sexual behavior (CSB) be considered as a behavioral addiction? A debate paper presenting the opposing view. J Behav Addict. (2022) 11:166–79. doi: 10.1556/2006.2020.00055 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Zeng X, Han X, Zheng D, Jiang P, Yuan Z. Similarity and difference in large-scale functional network alternations between behavioral addictions and substance use disorder: A comparative meta-analysis. Psychol Med. (2024) 54:473–87. doi: 10.1017/s0033291723003434 [DOI] [PubMed] [Google Scholar]
- 7. Liu Q-Q, Zhang D-J, Yang X-J, Zhang C-Y, Fan C-Y, Zhou Z-K. Perceived stress and mobile phone addiction in Chinese adolescents: A moderated mediation model. Comput Hum Behav. (2018) 87:247–53. doi: 10.1016/j.chb.2018.06.006 38826717 [DOI] [Google Scholar]
- 8. Lu GL, Ding YM, Zhang YM, Huang HT, Liang YP, Chen CR. The correlation between mobile phone addiction and coping style among Chinese adolescents: A meta-analysis. Child Adolesc Psychiatry Ment Health. (2021) 15:60. doi: 10.1186/s13034-021-00413-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Wang Z, Yang X, Zhang X. Relationships among boredom proneness, sensation seeking and smartphone addiction among Chinese college students: Mediating roles of pastime, flow experience and self-regulation. Technol Soc. (2020) 62:101319. doi: 10.1016/j.techsoc.2020.101319 38826717 [DOI] [Google Scholar]
- 10. Demirci K, Akgönül M, Akpinar A. Relationship of smartphone use severity with sleep quality, depression, and anxiety in university students. J Behav Addict. (2015) 4:85–92. doi: 10.1556/2006.4.2015.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Wang PW, Liu TL, Ko CH, Lin HC, Huang MF, Yeh YC, et al. Association between problematic cellular phone use and suicide: The moderating effect of family function and depression. Compr Psychiatry. (2014) 55:342–8. doi: 10.1016/j.comppsych.2013.09.006 [DOI] [PubMed] [Google Scholar]
- 12. Zhong Y, Ma H, Liang YF, Liao CJ, Zhang CC, Jiang WJ. Prevalence of smartphone addiction among Asian medical students: A meta-analysis of multinational observational studies. Int J Soc Psychiatry. (2022) 68:1171–83. doi: 10.1177/00207640221089535 [DOI] [PubMed] [Google Scholar]
- 13. Zenebe Y, Kunno K, Mekonnen M, Bewuket A, Birkie M, Necho M, et al. Prevalence and associated factors of internet addiction among undergraduate university students in Ethiopia: A community university-based cross-sectional study. BMC Psychol. (2021) 9:4. doi: 10.1186/s40359-020-00508-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Shen Y, Wang L, Huang C, Guo J, De Leon SA, Lu J, et al. Sex differences in prevalence, risk factors and clinical correlates of internet addiction among Chinese college students. J Affect Disord. (2021) 279:680–6. doi: 10.1016/j.jad.2020.10.054 [DOI] [PubMed] [Google Scholar]
- 15. Amano A, Ahmed G, Nigussie K, Asfaw H, Fekadu G, Hiko A, et al. Internet addiction and associated factors among undergraduate students of Jimma University; Jimma, South West Ethiopia, institutional based cross-sectional study. BMC Psychiatry. (2023) 23:721. doi: 10.1186/s12888-023-05197-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Bentéjac C, Csörgő A, Martínez-Muñoz G. A comparative analysis of gradient boosting algorithms. Artif Intell Rev. (2021) 54:1937–67. doi: 10.1007/s10462-020-09896-5 30311153 [DOI] [Google Scholar]
- 17. Cao S, Hu Y. Creating machine learning models that interpretably link systemic inflammatory index, sex steroid hormones, and dietary antioxidants to identify gout using the SHAP (SHapley Additive exPlanations) method. Front Immunol. (2024) 15:1367340. doi: 10.3389/fimmu.2024.1367340 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Kwon M, Kim D-J, Ch H, et al. The smartphone addiction scale: Development and validation of a short version for adolescents. PloS One. (2013) 8(12):e83558. doi: 10.1371/journal.pone.0083558 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Xiang M, Wang Z. Psychometric evaluation of the Chinese version of the smartphone dependence scale among adolescents. Chin. J. Clin. Psychol. 27:959–64. [Google Scholar]
- 20. Russell D, Peplau L, Cutrona C. The revised UCLA loneliness scale: Concurrent and discriminant validity evidence. J Pers Soc Psychol. (1980) 39:472–80. doi: 10.1037/0022-3514.39.3.472 [DOI] [PubMed] [Google Scholar]
- 21. Chiu JL, Wei LC. Loneliness as a predictor of disability in older adults: Implications of the UCLA loneliness scale cutoff score across cultural contexts. Int J Geriatric Psychiatry. (2024) 39(11):e70011. doi: 10.1002/gps.70011 [DOI] [PubMed] [Google Scholar]
- 22. Ogundimu EO, Altman DG, Collins GS. Adequate sample size for developing prediction models is not simply related to events per variable. J Clin Epidemiol. (2016) 76:175–82. doi: 10.1016/j.jclinepi.2016.02.031 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Wang Y, Abdullah MNLY. A meta-analytic review of smartphone addiction prevalence in Chinese college students. Curr Psychol. (2026) 45:548. doi: 10.1007/s12144-026-09160-z 30311153 [DOI] [Google Scholar]
- 24. Jiang Q, Li Y, Shypenka V. Loneliness, individualism, and smartphone addiction among international students in China. Cyberpsychol Behav Soc Netw. (2018) 21:711–8. doi: 10.1089/cyber.2018.0115 [DOI] [PubMed] [Google Scholar]
- 25. Kara N, Kara, Çetin M, Dönmez A, Kara M, Genç H. A study on the relationship between the levels of loneliness and smartphone addiction of students who are studying at the faculty of sports science. Asian J Educ Training. (2020) 6:213–8. doi: 10.20448/journal.522.2020.62.213.218 [DOI] [Google Scholar]
- 26. Cacioppo JT, Hawkley LC, Ernst JM, Burleson M, Berntson GG, Nouriani B, et al. Loneliness within a nomological net: An evolutionary perspective. J Res Pers. (2006) 40:1054–85. doi: 10.1016/j.jrp.2005.11.007 38826717 [DOI] [Google Scholar]
- 27. Kim Y-C. Media system dependency theory. In: The International Encyclopedia of Media Psychology, J. Bulck (Ed.). (2020), 1–17. doi: 10.1002/9781119011071.iemp0135 [DOI] [Google Scholar]
- 28. Chen Y, Xue Y, Lian S, Gu C, Dai H. Family socioeconomic status and smartphone dependence among college students: The mediating role of subjective well-being. Chin J Special Educ (Monthly). (2018), 80–4. doi: 10.3969/j.issn.1007-3728.2018.08.014 [DOI] [Google Scholar]
- 29. Gao D. Economic capital or cultural capital: Which matters more? The association between family background and consumption culture among university students. Soc Sci In Heilongjiang. (2015), 111–9. [Google Scholar]
- 30. Song A, Song G, Wang H, Niu Q, Yin G, Chen H, et al. Prevalence of mobile phone addiction among medical students: A systematic review. Am J Transl Res. (2023) 15:2985–98. [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.