A Multidimensional Dataset of Student Gaming Behavior and Academic Correlates in Bangladesh Authors: Md. Ismail Hossain, Lisan Al Amin, Minhaz Uddin Emon, Pooja Vishnu Chauhan, Akhil Rudrabatla, Faisal Quader Published: 2025 ABSTRACT In response to the growing interest in gaming behaviors and their societal implications, particularly in the Global South, we present a dataset documenting gaming habits among university students in Bangladesh. This initiative supports exploratory research on gaming engagement, including potential links to education, social dynamics, and self-reported well-being. The dataset captures behavioral patterns inspired by themes in the World Health Organization’s discussions on gaming health, while prioritizing cultural relevance for low-resource settings. By sharing this resource, we aim to advance understanding of gaming practices in socioeconomically diverse populations and encourage context-sensitive strategies to support balanced gaming habits. Researchers are cautioned that this dataset does not include clinical diagnostic tools for Internet Gaming Disorder (IGD) and should not be used to infer pathology. Background & Summary IGD has become a significant issue worldwide, particularly impacting the youth in the Global South, where its consequences on mental health, education, and social interactions are increasingly concerning. Bangladesh, as a representative of the Global South, experiences unique socio-economic challenges that exacerbate the effects of IGD among its university students1,2. While existing research has connected gaming habits with academic performance, there is a noted deficiency in exploring the broader implications of IGD, including its effects on physical health and social dynamics within this context3. Moreover, the lack of comprehensive datasets hampers the understanding of IGD’s multifaceted impact. Another issue is that the existing datasets2,4 are not sufficiently informative, which means that we are unable to benefit from the advancements in AI technology fully. Previously, Ali et al.5 investigated the psychometric properties of the Internet Gaming Disorder Scale (IGDS9-SF) among Sri Lankan university students, finding that the scale exhibited good reliability, criterion validity, and measurement invariance across different countries including Sri Lanka, Turkey, Australia, and the USA. This study highlighted the scale’s unidimensional structure and its invariance across different groups, though strict invariance was not maintained. The results emphasized that males and online multiplayer gamers showed higher IGD levels compared to females and offline players. De Pasquale et al.6 explored the prevalence of IGD among Italian university students, examining its associations with dissociative phenomena. Their findings revealed a high incidence of IGD risk among college students, particularly among males. The study demonstrated a positive correlation between IGD risk and dissociative experiences, including depersonalization, derealization, and imaginative involvement. The authors called for preventative interventions by public institutions to address the spread of IGD among youth.Pearcy et al.7 developed and validated a new measure for IGD based on the DSM-5 criteria, named the Personal Internet Gaming Disorder Evaluation-9 (PIE-9). The PIE-9 was tested on both student and community samples, showing high internal consistency and test-retest reliability. The measure also demonstrated convergent and predictive validity, distinguishing between individuals who met the IGD criteria and those who did not. The distress and disability associated with IGD, as measured by PIE-9, were found to be comparable to other common DSM-5 disorders, suggesting its potential utility as a screening tool in clinical practice. A consolidated summary of these studies is presented in Table 1.Table 1Comparison of Gaming Study Attributes.AuthorsSample SizeDemographicsGaming BehaviorFamily & Social InteractionEmotional/Physical ResponseEducationalLifestyleSayed et al.21,125✓Addiction× ✓× × Hawi et al.4524✓Addiction× ✓✓× Ali et al.5322✓Addiction× ✓✓× De Pasquale et al.6221✓Addiction× ✓× × Pearcy et al.7404✓Addiction× ✓× × ours9989✓Comprehensive✓✓✓✓ Table 1Comparison of Gaming Study Attributes.AuthorsSample SizeDemographicsGaming BehaviorFamily & Social InteractionEmotional/Physical ResponseEducationalLifestyleSayed et al.21,125✓Addiction× ✓× × Hawi et al.4524✓Addiction× ✓✓× Ali et al.5322✓Addiction× ✓✓× De Pasquale et al.6221✓Addiction× ✓× × Pearcy et al.7404✓Addiction× ✓× × ours9989✓Comprehensive✓✓✓✓ Comparison of Gaming Study Attributes. This study aims to address gaps in current literature by using advanced analytical techniques and a robust dataset. It aligns with the World Health Organization’s diagnostic guidelines for gaming disorders to explore the broader impact of gaming practices.8. In this paper, we introduce a comprehensive dataset9 derived from a study conducted over March and April 2023, focusing on university students in Bangladesh. This dataset9 encapsulates a broad spectrum of information, including demographic details, academic performance, gaming behaviors, emotional responses to gaming, levels of fatigue, and the prevalence of eyeglass usage among participants. Its richness and diversity offer an unprecedented opportunity for in-depth analysis of gaming habits and their implications. By making this dataset9 available, we aim to fill existing gaps in the literature and provide a foundation for future research in the field. The alignment with global gender-based gaming trends, alongside insights into unique local social activities and health impacts, positions this dataset as a crucial resource for advancing our understanding of gaming’s multifaceted effects. Methods A structured methodology, outlined in Fig. 1, guided every step of the research process-from the initial development of the questionnaire to the final interpretation of data. This section details the core components of the study’s methodologyFig. 1The process began with the creation of a questionnaire, which was then taken by university students through an online survey. Following this, the data was cleaned, processed, interpreted, and suggested how the data could be used for scientific applications. The process began with the creation of a questionnaire, which was then taken by university students through an online survey. Following this, the data was cleaned, processed, interpreted, and suggested how the data could be used for scientific applications. Ethical Considerations This study was reviewed and approved by the Institutional Review Board (IRB) at the University of Maryland, Baltimore County under protocol number 1360 (Applicable FWA #: 00000069). The research complied with ethical principles outlined in the Belmont Report and the U.S. Department of Health and Human Services (HHS) regulations at 45 CFR Part 46. All participants provided informed consent after being informed about the study’s purpose, voluntary participation, and their right to withdraw at any time. Data were collected remotely through online forms, and no personally identifiable information was collected. Participants consented to the anonymized use of their data for research and public sharing on open-access platforms. Under Daffodil International University’s (DIU) institutional policies, this type of minimal-risk, remote, and anonymous research does not require local IRB review. The IoT Lab at the Department of Computer Science and Engineering, DIU (Bangladesh), provided general support in raising awareness about the study and helping ensure that communications with participants were culturally and linguistically appropriate. However, no direct intervention, recruitment, or data collection was conducted on site at the university, and no identifiable institutional or student data were used. Participant A total of 1,061 university students from nine universities across Bangladesh participated in this study, recruited through university email lists, social media platforms, and student forums. To ensure diversity and minimize sampling bias, a stratified sampling approach was applied across public and private universities in different regions of Bangladesh. The inclusion criteria of the participants required that they be enrolled university students who regularly played video games. Participation was entirely voluntary, with no financial or academic incentives provided. Informed consent was obtained before data collection, ensuring participants were fully aware of the study’s purpose, confidentiality measures, and their right to withdraw at any time. Academic metrics, including CGPA (Mean = 3.13, SD = 0.56) and Higher Secondary School results (Mean = 4.16, SD = 0.66), were assessed alongside psychological parameters such as stress relief through gaming (affirmative: 51.3%). Behavioral aspects, including the age of gaming initiation (Mean = 19.53, SD = 2.57) and platform preference (Mobile: 49.9%; PC: 47.7%), were explored, as well as lifestyle factors like sleep duration (Mean = 9.63, SD = 10.51). Data Collection Method The data collection process aimed to examine gaming behaviors and their self-reported correlations with academic and social experiences among Bangladeshi university students. Survey questions were loosely adapted from thematic elements of the DSM-5’s IGD criteria (e.g., emotional states related to gaming) but do not operationalize its nine diagnostic criteria. The draft questionnaire was reviewed by experts to refine clarity and cultural appropriateness. Notably, the DSM-5 framework was used as a general reference rather than a strict diagnostic tool, as the study focuses on behavioral patterns rather than clinical pathology. A diverse sample of 1061 university students who regularly play games participated in the survey, which was conducted from March to April 2023. The data was collected electronically, ensuring anonymity and informed consent, and extensive pilot testing was done to ensure reliability. To mitigate social desirability and response biases, the survey was anonymized, questions were randomized in order, and completion time was monitored to detect potential survey fatigue. Data Processing In this study, we conducted rigorous data cleaning and transformation to ensure the integrity and reliability of the dataset, which initially comprised responses from 1,061 university students in Bangladesh. First, we addressed outliers in numerical variables, specifically the time spent playing games and time spent with family and friends, using the interquartile range (IQR) method. This refinement reduced the dataset to 1,014 samples. Additionally, we enforced logical constraints by removing entries where higher secondary school results exceeded 5.0 and current CGPA values surpassed 4.0, eliminating 22 invalid responses. Next, we standardized time-related data by converting inconsistent sleep time formats into a uniform 24-hour representation. Furthermore, we removed ambiguous or unrecognized game titles to maintain data consistency. Lastly, to enhance the quality of categorical variables, we filtered game addiction preferences by retaining only frequently chosen first-choice responses (≥4 occurrences), consolidating them into a single representative column. Following these pre-processing steps, the dataset9 was refined to 989 high-quality samples, ensuring its suitability for analysis. data transformation, numerical variables were normalized using min-max scaling and standardized with z-score scaling. Categorical variables were encoded with label encoding and one-hot encoding to facilitate their use in machine-learning models. We engineered new features, such as “Gaming Intensity,” and created interaction features to capture additional insights. This section is intended solely for machine learning purposes. After processing the data, a glimpse of the frequencies of a few variables is shown in Fig. 2.Fig. 2A glimpse of the variables and their frequencies in the datasets. A glimpse of the variables and their frequencies in the datasets. Intellectual Rationale for Questionnaire Design The questionnaire was structured to holistically capture gaming behaviors, their contextual drivers, and perceived impacts on students’ lives. Variables were organized into six thematic categories, each addressing distinct dimensions of gaming engagement. While we summarize the key variables in Table 2, detailing their types, measurement scales, and observed distributions, below we detail the rationale for these categories and their constituent variables. Demographics Capturing data on age, gender, and educational position is essential for stratified analyses, enabling the identification of trends within specific subgroups10. Age: (Variable Type: Continuous) - Represents the participant’s age. This variable can include any positive number, such as 18, 25, or 30 years old.Gender: (Variable Type: Categorical) - Represents the participant’s gender. The options include “Male”, “Female”, or “Other”.Educational Position: (Variable Type: Categorical) - Represents the participant’s current educational level. This can range from “High School”, “College”, “University”, to “Postgraduate”. Age: (Variable Type: Continuous) - Represents the participant’s age. This variable can include any positive number, such as 18, 25, or 30 years old. Age: (Variable Type: Continuous) - Represents the participant’s age. This variable can include any positive number, such as 18, 25, or 30 years old. Gender: (Variable Type: Categorical) - Represents the participant’s gender. The options include “Male”, “Female”, or “Other”. Gender: (Variable Type: Categorical) - Represents the participant’s gender. The options include “Male”, “Female”, or “Other”. Educational Position: (Variable Type: Categorical) - Represents the participant’s current educational level. This can range from “High School”, “College”, “University”, to “Postgraduate”. Educational Position: (Variable Type: Categorical) - Represents the participant’s current educational level. This can range from “High School”, “College”, “University”, to “Postgraduate”. Gaming Behavior & Preferences Questions on when and how students began gaming and their platform preferences provide insights into the influence of early exposure and technological accessibility, crucial for understanding the evolution of the gaming landscape11. Age Started Gaming: (Variable Type: Continuous) - Indicates the age at which the participant began gaming. Values could range from a young age like 5 years old to adolescence or later, such as 15 years old.Gaming Platform: (Variable Type: Categorical) - Represents the type of platform used by the participant for gaming. The options could include “PC”, “Console”, “Mobile”, or other forms of gaming platforms. Age Started Gaming: (Variable Type: Continuous) - Indicates the age at which the participant began gaming. Values could range from a young age like 5 years old to adolescence or later, such as 15 years old. Age Started Gaming: (Variable Type: Continuous) - Indicates the age at which the participant began gaming. Values could range from a young age like 5 years old to adolescence or later, such as 15 years old. Gaming Platform: (Variable Type: Categorical) - Represents the type of platform used by the participant for gaming. The options could include “PC”, “Console”, “Mobile”, or other forms of gaming platforms. Gaming Platform: (Variable Type: Categorical) - Represents the type of platform used by the participant for gaming. The options could include “PC”, “Console”, “Mobile”, or other forms of gaming platforms. Family & Social Interaction Examining the balance between gaming and social interactions helps explore the interplay between digital and physical socialization, which has significant implications for mental well-being12. Family and Friend Time: (Variable Type: Continuous) - Reflects how much time the participant spends with family and friends. This is typically represented by a number, such as 10 or 20 hours per week. Family and Friend Time: (Variable Type: Continuous) - Reflects how much time the participant spends with family and friends. This is typically represented by a number, such as 10 or 20 hours per week. Family and Friend Time: (Variable Type: Continuous) - Reflects how much time the participant spends with family and friends. This is typically represented by a number, such as 10 or 20 hours per week. Emotional & Physical Response Investigating emotional responses, fatigue, and using games for stress relief ties directly into the discourse on mental and physical health, reflecting the association between gaming and well-being13. Feelings without Gaming: (Variable Type: Categorical) - Indicates how the participant feels when unable to play games. They could feel “Frustrated”, “Calm”, “Anxious”, or “Neutral”.Feeling Fatigue: (Variable Type: Categorical) - Describes whether the participant feels fatigued after gaming. The answers could be “Yes” or “No”.Gaming for Stress Relief: (Variable Type: Categorical) - Reflects whether the participant uses gaming to relieve stress. This is answered with “Yes” or “No”. Feelings without Gaming: (Variable Type: Categorical) - Indicates how the participant feels when unable to play games. They could feel “Frustrated”, “Calm”, “Anxious”, or “Neutral”. Feelings without Gaming: (Variable Type: Categorical) - Indicates how the participant feels when unable to play games. They could feel “Frustrated”, “Calm”, “Anxious”, or “Neutral”. Feeling Fatigue: (Variable Type: Categorical) - Describes whether the participant feels fatigued after gaming. The answers could be “Yes” or “No”. Feeling Fatigue: (Variable Type: Categorical) - Describes whether the participant feels fatigued after gaming. The answers could be “Yes” or “No”. Gaming for Stress Relief: (Variable Type: Categorical) - Reflects whether the participant uses gaming to relieve stress. This is answered with “Yes” or “No”. Gaming for Stress Relief: (Variable Type: Categorical) - Reflects whether the participant uses gaming to relieve stress. This is answered with “Yes” or “No”. Educational Attributes Linking gaming habits to academic performance addresses the debate on whether gaming hinders or aids academic pursuits, particularly in the high-pressure academic environment of Bangladesh14. Current CGPA: (Variable Type: Continuous) - Represents the participant’s current cumulative grade point average (CGPA). This is typically a decimal number, such as 3.5 or 4.0.Higher Education Result: (Variable Type: Categorical) - Represents the participant’s long-term academic result in higher education. The values could be “Excellent”, “Good”, “Average”, or “Poor”. Current CGPA: (Variable Type: Continuous) - Represents the participant’s current cumulative grade point average (CGPA). This is typically a decimal number, such as 3.5 or 4.0. Current CGPA: (Variable Type: Continuous) - Represents the participant’s current cumulative grade point average (CGPA). This is typically a decimal number, such as 3.5 or 4.0. Higher Education Result: (Variable Type: Categorical) - Represents the participant’s long-term academic result in higher education. The values could be “Excellent”, “Good”, “Average”, or “Poor”. Higher Education Result: (Variable Type: Categorical) - Represents the participant’s long-term academic result in higher education. The values could be “Excellent”, “Good”, “Average”, or “Poor”. Lifestyle Factors: Questions on sleep patterns, newspaper reading habits, and class attendance provide a broader view of students’ daily routines and how gaming fits into or disrupts these routines15. Sleep Time: (Variable Type: Continuous) - Describes how many hours the participant sleeps, such as 6 or 8 hours each night.Newspaper Reading: (Variable Type: Categorical) - Indicates whether the participant reads newspapers. They could answer “Yes” or “No”.Morning Class Attendance: (Variable Type: Categorical) - Shows if the participant attends morning classes regularly, with possible responses of “Yes” or “No”.Wearing Glasses: (Variable Type: Categorical) - Represents if the participant wears glasses, with “Yes” or “No” as the potential answers. Sleep Time: (Variable Type: Continuous) - Describes how many hours the participant sleeps, such as 6 or 8 hours each night. Sleep Time: (Variable Type: Continuous) - Describes how many hours the participant sleeps, such as 6 or 8 hours each night. Newspaper Reading: (Variable Type: Categorical) - Indicates whether the participant reads newspapers. They could answer “Yes” or “No”. Newspaper Reading: (Variable Type: Categorical) - Indicates whether the participant reads newspapers. They could answer “Yes” or “No”. Morning Class Attendance: (Variable Type: Categorical) - Shows if the participant attends morning classes regularly, with possible responses of “Yes” or “No”. Morning Class Attendance: (Variable Type: Categorical) - Shows if the participant attends morning classes regularly, with possible responses of “Yes” or “No”. Wearing Glasses: (Variable Type: Categorical) - Represents if the participant wears glasses, with “Yes” or “No” as the potential answers. Wearing Glasses: (Variable Type: Categorical) - Represents if the participant wears glasses, with “Yes” or “No” as the potential answers. Data Records The dataset is available at Zenodo9, with this section being the primary source of information on the availability. The data used in this study are also available online and include both the training and test sets hosted on Hugging Face.igd_data.csv: This cleaned dataset contains 989 entries from university students in Bangladesh who regularly play video games. It includes 18 variables: Age, Current CGPA, H. SC Result, Gaming Start Age, Sleep Time, Average Gaming Time, Family Time, Gender, Educational Level, Gaming Platform, Morning Class Attendance, Newspaper Reading, Game Type Preference, Use of Gaming for Stress Relief, Game Addiction Type, Feelings on a Day without Gaming, and Use of Eyewear. igd_data.csv: This cleaned dataset contains 989 entries from university students in Bangladesh who regularly play video games. It includes 18 variables: Age, Current CGPA, H. SC Result, Gaming Start Age, Sleep Time, Average Gaming Time, Family Time, Gender, Educational Level, Gaming Platform, Morning Class Attendance, Newspaper Reading, Game Type Preference, Use of Gaming for Stress Relief, Game Addiction Type, Feelings on a Day without Gaming, and Use of Eyewear. igd_data.csv: This cleaned dataset contains 989 entries from university students in Bangladesh who regularly play video games. It includes 18 variables: Age, Current CGPA, H. SC Result, Gaming Start Age, Sleep Time, Average Gaming Time, Family Time, Gender, Educational Level, Gaming Platform, Morning Class Attendance, Newspaper Reading, Game Type Preference, Use of Gaming for Stress Relief, Game Addiction Type, Feelings on a Day without Gaming, and Use of Eyewear. The original dataset with all 1,061 responses is available in the igd_responses_raw.csv file, while the igd_data.csv file reflects the cleaned version with 989 valid samples. These files are freely accessible via GitHub (https://github.com/ismail31416/IGD) and Hugging Face (https://huggingface.co/datasets/ismail31415/uniGame). The repository also includes a README file providing detailed descriptions of the dataset, preprocessing steps, and relevant metadata. The primary source of the dataset can be accessed via Zenodo at 10.5281/zenodo.15068893. Technical Validation In this section, we present a comprehensive technical validation of the dataset, including comparisons with prior studies, statistical consistency checks, and a discussion of its limitations. Data Comparison The cited studies (Table 1) explore IGD across diverse populations but differ in scope, methodology, and limitations. Sayed et al.2 focused on mobile game addiction among Bangladeshi university students, identifying prevalence and sociodemographic factors but lacked clinical validation. Hawi et al.4 examined Lebanese students, linking IGD to poor sleep and academic performance, yet their cross-sectional design limited causal inferences. De Pasquale et al.6 studied Italian students, associating IGD with dissociative symptoms, but their small sample (N=221) reduced generalizability. Ali et al.5 validated the IGDS9-SF scale across four countries (Sri Lanka, Turkey, Australia, USA), demonstrating cross-cultural reliability but omitting behavioral or academic correlates. Pearcy et al.7 introduced the PIE-9 scale, showing strong psychometric properties but focused narrowly on diagnostic criteria without contextual data (e.g., lifestyle or education). In contrast, the current dataset9 (Bangladesh, N=989) offers a holistic view by integrating gaming behavior with academic metrics (CGPA), emotional responses (stress relief), and lifestyle factors (sleep, social interaction), addressing gaps in prior studies. However, it avoids clinical IGD diagnosis, limiting pathological inferences. While earlier works prioritized scale validation (Ali, Pearcy) or singular outcomes (Hawi: sleep; De Pasquale: dissociation), this study’s multidimensional approach enables broader socio-academic analyses, albeit within a single geographic context.Table 2Variable Summary: Modality, Scale, and Observed Values.Variable (Type)Modality/ScaleObserved Values (Frequency/Stats)Age (Cont.)ContinuousMean = 22.99, SD = 3.16Gender (Cat.)Categorical (Nominal)Male (50.4%), Female (49.6%)Educational Position (Ord.)Ordinal (Bachelor < Master’s)Bachelor (68.6%), Master’s (31.4%)Gaming Platform (Cat.)Categorical (Nominal)Mobile (49.9%), PC (47.7%), Both (2.3%)Average Gaming Time (Cont.)Continuous (0-8 hrs/day)Mean = 3.77, SD = 1.59Feelings w/o Gaming (Likert)Likert Scale (1 = Negative, 2 = Neutral, 3 = Positive)Negative (73.0%), Positive (26.2%), Neutral (0.7%)Gaming for Stress Relief (Dich.)Dichotomous (Yes/No)Yes (51.3%), No (48.7%)Current CGPA (Cont.)Continuous (2.00-4.00)Mean = 3.13, SD = 0.56Sleep Time (Cont.)Continuous (0-23 hrs)Mean = 9.63, SD = 10.51Class Attendance (Dich.)Dichotomous (Regular/Not Regular)Regular (50.9%), Not Regular (49.1%)Wearing Glasses (Dich.)Dichotomous (Yes/No)Yes (45.9%), No (54.1%)Game Addiction Type (Cat.)Categorical (Nominal)PUBG (15.8%), Free Fire (15.3%), CoD (14.1%), IGA (14.1%), PES (13.1%), FIFA (13.6%), COC(13.0%) Table 2Variable Summary: Modality, Scale, and Observed Values.Variable (Type)Modality/ScaleObserved Values (Frequency/Stats)Age (Cont.)ContinuousMean = 22.99, SD = 3.16Gender (Cat.)Categorical (Nominal)Male (50.4%), Female (49.6%)Educational Position (Ord.)Ordinal (Bachelor < Master’s)Bachelor (68.6%), Master’s (31.4%)Gaming Platform (Cat.)Categorical (Nominal)Mobile (49.9%), PC (47.7%), Both (2.3%)Average Gaming Time (Cont.)Continuous (0-8 hrs/day)Mean = 3.77, SD = 1.59Feelings w/o Gaming (Likert)Likert Scale (1 = Negative, 2 = Neutral, 3 = Positive)Negative (73.0%), Positive (26.2%), Neutral (0.7%)Gaming for Stress Relief (Dich.)Dichotomous (Yes/No)Yes (51.3%), No (48.7%)Current CGPA (Cont.)Continuous (2.00-4.00)Mean = 3.13, SD = 0.56Sleep Time (Cont.)Continuous (0-23 hrs)Mean = 9.63, SD = 10.51Class Attendance (Dich.)Dichotomous (Regular/Not Regular)Regular (50.9%), Not Regular (49.1%)Wearing Glasses (Dich.)Dichotomous (Yes/No)Yes (45.9%), No (54.1%)Game Addiction Type (Cat.)Categorical (Nominal)PUBG (15.8%), Free Fire (15.3%), CoD (14.1%), IGA (14.1%), PES (13.1%), FIFA (13.6%), COC(13.0%) Variable Summary: Modality, Scale, and Observed Values. Data Quality Check To ensure the reliability and integrity of the dataset9, we conducted rigorous statistical checks and machine learning validation. These steps were designed to confirm the absence of noise, validate variable distributions, and assess the dataset’s suitability for downstream analyses. Statistical Checks Statistical tests were employed to verify the meaningfulness of variable relationships and identify potential anomalies. The Chi-Squared test revealed statistically significant associations (p < 0.05) between key variables such as gaming platform preference and stress relief behavior (χ2 = 15.2, p = 0.004), confirming that observed patterns were unlikely to occur by chance (Fig. 3). Variables exhibited expected distributions: gender maintained near-balanced proportions (50.4% male vs. 49.6% female), while educational positions followed institutional enrollment trends (68.6% bachelor’s vs. 31.4% master’s). Continuous variables showed plausible ranges - average gaming time (μ = 3.77 hrs/day, σ = 1.59) aligned with typical student schedules, and CGPA values (μ = 3.13/4.00) reflected academic norms. Sleep time variability (σ = 10.51 hrs) was investigated for potential outliers, but manual verification confirmed valid entries reflecting diverse student schedules.Fig. 3Heatmap showing statistically significant relationships among key variables, used to assess internal consistency and validate the reliability of the dataset. Heatmap showing statistically significant relationships among key variables, used to assess internal consistency and validate the reliability of the dataset. Machine Learning Validation We established predictive validity through supervised learning benchmarks targeting the stress relief variable. XGBoost achieved 91.92% accuracy (F1-score = 92.66%), demonstrating the dataset’s capacity to support robust predictive modeling. The fairness evaluation using disparate impact scores (DIS)16 revealed moderate bias (53.81% for XGBoost), prompting stratification analyses that confirmed consistent performance across gender subgroups (Δaccuracy < 2%). Ten diverse algorithms17–20 were tested (Table 3), with tree-based models outperforming linear methods (91.41-91.92% vs. 58.08-73.23% accuracy), suggesting non-linear relationships that match the complexity of gaming behavior phenomena.Table 3Comparison of Various Machine Learning Models for Predicting “Do you play games for stress relief?”ModelAccuracy (%)Precision (%)Recall (%)F1-Score (%)ROC-AUC (%)DIS(%)LogReg58.0863.2160.3661.7559.440.00k-NN63.1368.2763.9666.0568.370.00SVM73.2377.3673.8775.5882.780.00DecTree82.3285.8581.9883.8781.0253.30RandForest91.4197.0087.3991.9498.6150.25GradBoost71.7278.3568.4773.0881.240.00AdaBoost56.5762.3856.7659.4362.710.00XGBoost91.9294.3990.9992.6697.7353.81LightGBM88.8994.9584.6889.5297.3849.75NeuralNet85.8686.2085.0286.0089.4049.23 Table 3Comparison of Various Machine Learning Models for Predicting “Do you play games for stress relief?”ModelAccuracy (%)Precision (%)Recall (%)F1-Score (%)ROC-AUC (%)DIS(%)LogReg58.0863.2160.3661.7559.440.00k-NN63.1368.2763.9666.0568.370.00SVM73.2377.3673.8775.5882.780.00DecTree82.3285.8581.9883.8781.0253.30RandForest91.4197.0087.3991.9498.6150.25GradBoost71.7278.3568.4773.0881.240.00AdaBoost56.5762.3856.7659.4362.710.00XGBoost91.9294.3990.9992.6697.7353.81LightGBM88.8994.9584.6889.5297.3849.75NeuralNet85.8686.2085.0286.0089.4049.23 Comparison of Various Machine Learning Models for Predicting “Do you play games for stress relief?” Limitation While this dataset9 provides valuable insights, several limitations should be noted. First, self-reported measures may introduce response bias, particularly for academic performance and gaming habits. Second, the absence of clinical diagnostic tools for IGD restricts direct pathological inferences. Third, the cross-sectional design limits causal interpretations between gaming behaviors and outcomes. Finally, the focus on Bangladeshi university students may reduce generalizability to other populations. These constraints highlight opportunities for future work, including longitudinal designs and integration of validated IGD assessments. Usage Notes This dataset, comprising various demographic, lifestyle, emotional, and educational attributes related to gamers, is intended for researchers, policymakers, and developers in psychology, game development, and social sciences. The data9 are available at an institutional data mirror for convenience: 10.5281/zenodo.15068893 Software Packages Analyzing the dataset9 can be performed using standard statistical software packages such as R, SPSS, or Python’s SciPy and pandas libraries. Visualization can be facilitated through ggplot2 in R or Matplotlib in Python. Processing Steps Downstream processing may include normalization, especially for continuous variables like average gaming time or CGPA, to allow comparisons across different scales. Outliers handling, missing value imputation, and categorization of continuous variables may also be required. Clustering or segmentation techniques might be helpful to identify patterns or segments within the dataset. Integration with Other Datasets Researchers looking to compare or integrate this dataset with others should ensure consistency in terms like measurement scales, definitions, and variable categories. Harmonizing data would enhance comparability and provide a richer understanding of gaming and its multifaceted impacts. Privacy Controls Access to the complete dataset is subject to privacy restrictions due to the inclusion of personal and potentially sensitive information. Researchers interested in accessing the full dataset must apply through the designated data access committee. Criteria for access include adherence to ethical guidelines, a clear research proposal, and compliance with all relevant laws and regulations. Limitations on data use may include anonymization, restricted dissemination, or other measures as stipulated by relevant legal and ethical standards. Researchers are encouraged to contact the corresponding author for further details on specific attributes, methodologies, or data access procedures.