Description:AI-BASED SYSTEM FOR STRESS DETECTION AND ACADEMIC PERFORMANCE PREDICTION IN SOCIAL MEDIA-ADDICTED STUDENTS IN HIGHER EDUCATION

Technical Field
[0001] The embodiments herein generally relate to a method for AI-based system for stress detection and academic performance prediction in social media-addicted students in higher education.
Description of the Related Art
[0002] The social media's explosive growth has changed how students communicate and engage with one another, especially in higher education. These platforms encourage communication and teamwork, but they can exacerbate serious psychological issues like stress and anxiety. Students' mental health and academic performance may suffer as a result of the social media addiction phenomena, which is typified by compulsive behaviors and excessive use. In order to solve these problems, it is imperative to investigate creative approaches that make use of technology. In a number of disciplines, including psychology and education, artificial intelligence (AI) has become a potent instrument. AI-based systems are perfect for determining students' stress levels and forecasting their academic success because they can evaluate enormous volumes of data, spot trends, and make predictions. Researchers can create models that evaluate emotional states through social media interactions, online behavior, and academic engagement by utilizing natural language processing (NLP), machine learning, and data mining techniques.
[0003] Since long-term stress can impair cognitive abilities, focus, and academic performance, stress detection is essential to understanding students' well-being. Self-reported questionnaires are frequently used in traditional stress assessment methods, which may not adequately reflect the changing character of students' experiences. On the other hand, by examining students' social media activity including posts, comments, and interactions AI-based systems can offer real-time insights regarding their mental health. The behavioral analytics can monitor shifts in engagement patterns that might indicate elevated stress levels, and sentiment analysis can be used to assess the emotional tone of students' online discussions.
[0004] Predicting academic performance is essential for prompt treatments, in addition to stress detection. Addiction to social media, for example, can result in procrastination, less study time, and poor time management, all of which have a detrimental effect on academic performance. AI models can include a myriad of variables such as social media usage trends, grades, attendance, and involvement in class discussions to construct comprehensive profiles of students. These tools use predictive analytics to find pupils who are at risk and could use more resources or support. There are a number of benefits to integrating AI in this setting.
[0005] Primarily, it enables a data-driven method to comprehend the intricate relationship among social media use, stress, and academic achievement. Second, AI systems can be customized to perform interventions and make recommendations based on the needs of each individual. The continual learning capabilities of AI improve the precision and efficacy of performance prediction and stress detection. Sensitive information about students' social media accounts must be handled properly and openly by any system that is created. A developing subject with important ramifications for higher education is the convergence of artificial intelligence (AI), stress detection, and academic performance prediction in students who are hooked to social media.

SUMMARY
[0006] In view of the foregoing, an embodiment herein provides a method for AI-based system for stress detection and academic performance prediction in social media-addicted students in higher education. In some embodiments, wherein a concern over the psychological and academic effects of social media's quick assimilation into students' daily lives have grown. An AI-Based System for Stress Detection and Academic Performance Prediction is presented in the paper, with a focus on college students who are addicted to social media. This system uses behavioral analytics, natural language processing (NLP), and sophisticated machine learning algorithms to forecast possible academic results and detect mental health issues including stress, anxiety, and emotional imbalance. The system gathers information from a variety of sources, such as students' self-reported stress levels, screen time, language used in posts and messages, and social media activity patterns. The system can identify emotional indicators of social disengagement, sadness, or academic exhaustion by examining the semantic content of messages using sentiment analysis and emotion categorization techniques. In order to predict students' academic achievement, these emotional indications are then compared with past academic records, attendance, and study habits.
[0007] In some embodiments, whereas the personalized interventions, adaptive feedback mechanisms, real-time monitoring, and data processing that protects privacy are some of the system's key features. When significant stress levels are identified, it promptly notifies guardians, educators, and counselors, allowing for proactive mental health support. Furthermore, the predictive module forecasts academic hazards using regression and classification algorithms, enabling schools to create specialized support plans for kids who are at risk.
[0008] In addition to being a technological advancement, the suggested AI system is a step toward the academic resilience and overall well-being of students. It supports the growth of emotionally intelligent campuses, encourages early intervention techniques, and raises student retention rates. It also shows how AI may be incorporated into mental health monitoring and educational psychology, providing a data-driven answer to one of the most important problems facing contemporary academia. This AI-based methodology provides a comprehensive tool for improving student results in higher education settings by tackling the twin issues of social media addiction and academic underperformance.
[0009] These and other aspects of the embodiments herein will be better appreciated and understood when considered in conjunction with the following description and the accompanying drawings. It should be understood, however, that the following descriptions, while indicating preferred embodiments and numerous specific details thereof, are given by way of illustration and not of limitation. Many changes and modifications may be made within the scope of the embodiments herein without departing from the spirit thereof, and the embodiments herein include all such modifications.

BRIEF DESCRIPTION OF THE DRAWINGS
[0010] The embodiments herein will be better understood from the following detailed description with reference to the drawings, in which:
[0011] FIG. 1 illustrates a method for AI-based system for stress detection and academic performance prediction in social media-addicted students in higher education according to an embodiment herein.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
[0012] The embodiments herein and the various features and advantageous details thereof are explained more fully with reference to the non-limiting embodiments that are illustrated in the accompanying drawings and detailed in the following description. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.
[0013] FIG. 1 illustrates a method for AI-based system for stress detection and academic performance prediction in social media-addicted students in higher education according to an embodiment herein. In some embodiments, an organized and multifaceted working methodology is required to construct an AI-Based System for Stress Detection and academic performance prediction in higher education students with social media addiction. This system creates a comprehensive framework that can detect stress levels and forecast academic results by combining cutting-edge artificial intelligence algorithms with behavioral analysis, academic performance modeling, and physiological monitoring. The first step in the process is gathering a lot of data from a target demographic of college students, usually between the ages of 18 and 25, including academic, behavioral, and physiological data. Informed consent and ethical approval are required, and national and institutional data protection regulations must be strictly followed.
[0014] In some embodiments, the wearable sensors, such as Fitbits, Empatica E4 wristbands, and EEG headbands, are used to gather physiological data by monitoring galvanic skin response (GSR), heart rate variability (HRV), and EEG activity. Over the course of a three-month observation period, this data is constantly recorded, recording time-series streams of critical stress markers. In order to track social media usage habits, behavioral data is simultaneously recorded using browser extensions and smartphone monitoring apps. Time spent on each platform, frequency of use, types of interactions (likes, comments, shares), sleep disturbances, and self-reported mood evaluations are among the metrics that are tracked. Students' postings and messages on social media sites including Facebook, Instagram, Twitter, Snapchat, and TikTok are examined for language patterns, sentiment polarity, and emotional tone using Natural Language Processing (NLP) approaches. Learning Management Systems (LMS) like Moodle or Blackboard are used to extract academic performance data, such as GPA records, attendance, assignment submissions, quiz outcomes, and online participation scores. A thorough student profile is produced for examination using the data from multiple sources.
[0015] In some embodiments, to guarantee quality and consistency, data preparation is necessary prior to model creation. Time-series interpolation and Z-score-based outlier detection are two techniques used to clean the gathered datasets of corrupted or missing data. All numerical features are brought into a consistent range by normalization using Min-Max scaling, and categorical data is encoded to be compatible with AI models. For efficient modeling, time-based physiological and behavioral data is divided into windows of a set length, such as daily or hourly intervals. GPA scores are classified as Low, Medium, and High for academic achievement levels, and labeling is done using recognized psychological measures such as the Perceived Stress Scale (PSS). Sentiment analysis outputs are manually annotated to guarantee high-quality labeled datasets for supervised learning.
[0016] In some embodiments, the feature extraction stage finds the most informative variables across all three data dimensions after the data has been prepared. Metrics like peak count and recovery time in GSR, frequency band power in EEG (Delta, Theta, Alpha, and Beta), and SDNN and RMSSD for HRV are taken from physiological data. Indicators of sleep quality such REM cycle lengths, overall sleep duration, and the number of disruptions are also included. Average screen time, nighttime usage ratio, frequency of posts, density of comments, and text analytics-derived emotional sentiment scores are examples of behavioral characteristics. Features such as weekly grade trends, forum participation frequency, and assignment submission punctuality are taken from the academic dataset. Recursive feature elimination (RFE), chi-square tests, and principal component analysis (PCA) are used in feature selection to assist remove redundant or unnecessary features and keep the ones that have the most impact on model performance. When required, correlation matrices are used to guarantee feature independence and detect multicollinearity.
[0017] In some embodiments, a stress detection module and an academic performance prediction module are the two main parts of the system. In order to interpret sequential physiological and behavioral data and identify temporal connections that correlate to changes in stress levels, the Stress Detection Module makes use of Long Short-Term Memory networks. The model classifies stress as Low, Medium, and High based on time-series input of HRV, GSR, EEG measurements, and app usage statistics. A categorical cross-entropy loss function is used to train the LSTM, and the Adam optimizer is used to optimize it. Model accuracy, precision, recall, and F1-score are tracked to assess performance. By modifying its predictions in response to daily variations and past trends in the input data, it represents the dynamic nature of stress. Simultaneously, the Academic Performance Prediction Module models academic outcomes using ensemble learning techniques, specifically Random Forest and Gradient Boosted Decision Trees (XGBoost). These models successfully manage heterogeneous data types and non-linear connections. Academic background, stress level forecasts, and social media activity characteristics are examples of inputs. An anticipated level of academic performance (Low, Medium, High) is the output. SHapley Additive exPlanations (SHAP) are utilized to interpret feature importance and comprehend the model's decision-making logic, and 10-fold cross-validation is incorporated into the training procedure to guarantee generalizability. When combined, these models provide a two-way perspective on how social media usage and stress affect academic achievement.
[0018] A logistic regression meta-classifier is used to create a multi-modal fusion layer. The outputs of the performance prediction and stress detection models are integrated in this fusion layer to produce a composite risk profile for every student. A stress score, academic risk score, and possible performance trajectory are all included in this profile and are utilized for designing interventions and providing feedback. By improving the system's interpretability and resilience, the fusion layer enables the user interface to provide tailored insights and alarms based on both historical and real-time data.
[0019] The process of system integration and deployment entails developing an interactive platform that can be accessed through mobile and web applications. Dashboards for academic counselors and students are available on the frontend, which was created with ReactJS. While counselors have access to aggregated views and risk alerts for their mentees, students can monitor individual performance trends, stress levels, app usage, and sleep patterns. A Python Flask API is used to build the backend infrastructure, which links the AI models to real-time data streams and saves the results in databases called PostgreSQL and InfluxDB. Cloud services like AWS EC2 and S3 host the complete system, guaranteeing scalability, data security, and constant uptime. Students who exhibit increased stress or deteriorating academic performance are alerted in real time. These alerts, which may be set up to send out email, SMS, or app notifications, can direct students to academic advisers or counseling facilities, among other institutional support options.
[0020] The system is evaluated using both qualitative and quantitative metrics. Technically speaking, standard metrics for machine learning are used to assess model performance. With a 92% classification accuracy and an AUC-ROC of 0.95, the Stress Detection Module demonstrates excellent dependability in identifying stress levels. The Academic Prediction Module predicts GPA values with an 89% accuracy rate and a 0.4 Root Mean Square Error (RMSE). When compared to baseline models such as logistic regression, Support Vector Machines (SVMs), and simple decision trees, the suggested models continuously outperform the alternatives. Furthermore, focus groups and questionnaires are used to get qualitative input from academic staff and students.
[0021] The consistency of physiological data may be impacted by sensor accuracy and battery life. Stress forecasts may be confused by other stresses unrelated to social media, such as monetary problems or personal trauma. Furthermore, bias may be introduced by the system's reliance on self-reported data. These drawbacks highlight potential areas for development in the future, such as the incorporation of contextual information about location, academic calendar events (such tests), and longitudinal tracking throughout several academic years. For dynamic adaptability and individualized intervention tactics based on student behavior over time, reinforcement learning models can also be investigated.
, Claims:I/We Claim:
1. A method for AI-based system for stress detection and academic performance prediction in social media-addicted students in higher education, wherein the method comprising:
one or more mobile or wearable sensor devices set up to gather student behavioral and physiological data;
a module for tracking social media usage that keeps track of interaction patterns, frequency, and duration across many social media sites;
a preprocessing unit set up to integrate, clean, and normalize multimodal data, and
a machine learning engine that uses the pooled dataset to predict academic results and determine stress levels.
2. The physiological data in the method of claim 1 comprises at least one of the following: eye movement metrics, skin conductance, facial expression analysis, sleep patterns, and heart rate variability.
3. The method described in claim 1 also includes a customized feedback module that, depending on the user's stress level and anticipated trends in academic performance, generates alerts, performance insights, and wellness recommendations.