MACHINE LEARNING-BASED APPROACHES FOR PREDICTING STUDENT STRESS IN HIGHER EDUCATION

Abstract

The method for the development of using the machine learning technique, an ensemble prediction model has been proposed to determine the academic stress of various students studying in various disciplines. In order to help students manage their anxiety, stress, and depression, this study attempted to identify the students' academic stress early on. This allowed their parents, teachers, college administration, and others to take various preventative measures. The dataset generated from the questionnaire is used to experimentally validate the suggested framework. To determine which parameters are most important to the students, we also examine their responses. Therefore, in order to take corrective action, our proposed work can be used to determine the mental stress level of students. In order to further characterize understudies, the model establishes a two-way arrangement of anxiety, determining whether the understudy is peaceful or unpleasant. It also determines whether the understudy's pressure rates are low, medium, or high. FIG.1

Specification

Description:MACHINE LEARNING-BASED APPROACHES FOR PREDICTING STUDENT STRESS IN HIGHER EDUCATION

Technical Field
[0001] The embodiments herein generally relate to a method for machine learning-based approaches for predicting student stress in higher education.
Description of the Related Art
[0002] The sudden explosion of the COVID-19 virus, which caused a pandemic, the process of digitizing the teaching and learning process has accelerated. The majority of nations in the world adopted and rigorously followed social distancing and work from home policies in an effort to stop the Corona virus from spreading and to slow its growth curve. When the outbreak happened, kindergartens, schools, colleges, universities, and other similar institutions were closed. The education sector is not an exception to this rule. Globally, mental illnesses are very common and significantly increase morbidity and mortality. According to a 2019 study, 6.9 million Bangladeshis, or 4.4% of the country's total population, suffer from anxiety. Not only will the stress they experience during their schooling affect their working environments, but it will also negatively impact their personal lives, health, and IQ. According to a recent research survey, these competitive curricula will result in more unhealthy future generations unless universities and other institutions pay attention when creating them with the right stress management additives.
[0003] The students' anxiety and stress levels are raised by the current online learning environment and the fierce global competition. Peer pressure, parental pressure, health problems, and financial circumstances are additional factors that lead to mental disparities among students. In particular, students' perceived stress has increased due to peer and parental pressure, frequent network outages, isolation, and taking classes online. The coronavirus pandemic has been an additional factor, disrupting students' daily routines and making them feel under additional pressure, which has resulted in subpar performance. Stress is the response to any external stimulus that is interpreted as threatening or dangerous. 50% of people worldwide suffer from this mental state, which affects their physical and mental well-being. This is because it is an organism's adaptive response to its surroundings, resulting in emotional, behavioral, and cognitive changes that have an impact on the health of those who experience it. Students experience anxiety and stress as a result of the current educational system and the intense competition. Students' mental inequality is caused by a variety of factors, including peer pressure, parental pressure, health issues, and socioeconomic status.
[0004] Very little automation has been used in educational institutions and organizations to predict student stress. One of the most important tasks is to observe each student and their profile. This duty falls under human interaction, which is why the suggested work opens the door for the automatic stress prediction of each student's failure under different conditions and appropriately suggests a solution for each student. Given the significance of comprehending the elements that influence student learning, the prediction of students' stress and anxiety has drawn the interest of numerous researchers worldwide. As a result, machine learning techniques are among the most popular ways to make predictions. Even though the new idea may be simpler or more complex, society will immediately reject it if it is supposed to be adopted by the established one. The key mantra is to give the new idea enough time to be accepted by the students in order to prevent such rejection. Students' mindsets will naturally become stressed if they are not given enough attention and time flexibility.
SUMMARY
[0001] In view of the foregoing, an embodiment herein provides a method for machine learning-based approaches for predicting student stress in higher education. In some embodiments, wherein the study could be enhanced by adding more machine learning algorithms, expanding the dataset, and using feature engineering techniques to add some new features. Seldom has the Crisp-DM methodology been applied to forecast students' stress and anxiety. As a result, they employ a methodology in other works that is based on a series of steps that typically include acquiring the dataset, the tools to be used and how to use them, the technological platform, the algorithms, and the outcomes. This methodology was found in more studies that predicted stress than those that dealt with anxiety. The authors used data from the 2017 OSMI open sourcing mental illness survey in the IT sector. This dataset, which contains labels like gender, age, and family history, has been subjected to a range of machine learning algorithms. Their findings indicate that 75% of people in the IT industry are vulnerable to pressure.
[0002] In some embodiments, wherein the current study's limitations, such as the quarantine period and movement restrictions, made it difficult to get in touch with students and gather data in an organized manner. The second restriction concerned the use of online Google Forms for data collection, which prevented participation in the study by students without internet access. According to studies, this occurs because female students are not afforded the same opportunities as male students. While women develop similarly to men in college, they are still observed to perform 60% of household duties at home, compared to 40% for men. Similarly, it's clear that some professions, like engineering, are still associated with men. Emotional, physical, and depressive symptoms are the most common psychological factor associated with anxiety. It is also possible to determine that the characteristics of stress that most stood out were feeling extremely exhausted or worn out and experiencing emotional, physical, or depressive symptoms. These were used to measure or record the automatic stress detection. The value of the index, which served as a baseline for stress prediction, was compared to the data gathered from all external and internal sensors.
[0003] In some embodiments, wherein by taking into account the example of a university, the suggested work can be expanded further. To find out how well their students are doing, colleges should take the stress prediction test after a few months. We can learn from this survey that students are studying well and are not experiencing excessive stress. The factors that contributed to the highest percentage of anxiety were family economy, monthly income, and financial situation. However, when it comes to stress, the characteristics that are most noticeable are: the parents' occupation, the relationship with the family or challenging environment, and the financial status, monthly income, or family economy. A survey was conducted across a number of industries. It focused on three areas: the physical, environmental, and psychological. The analysis employs both metric and non-metric approaches.
[0004] 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
[0001] The embodiments herein will be better understood from the following detailed description with reference to the drawings, in which:
[0002] FIG. 1 illustrates a method for machine learning-based approaches for predicting student stress in higher education according to an embodiment herein; and
[0003] FIG. 2 illustrates a method for database search criteria according to an embodiment herein.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
[0001] 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.
[0002] FIG. 1 illustrates a method for machine learning-based approaches for predicting student stress in higher education according to an embodiment herein. In some embodiments, the Google form is the most convenient way to reach responses, such as those from graduated students, it is used to gather data. Personal information, educational information, and other information are among the three primary question types covered in the online survey. The classification process made use of the 560 responses that were gathered. Seventy-five percent of the sample came from different engineering disciplines, while the remaining portion came from medical disciplines. The methods of knowledge collection made a significant contribution to our prediction model. Since it affects how stressed-out university students appear, the aforementioned factor is most frequently used in research. The way our "alert system" responds to changes can therefore be considered a reflection of stress. According to studies conducted on college students, confinement, isolation, and lifestyle modifications had a greater impact on their mental health and the conditions under which students were expected to live. An Excel spreadsheet was used to collect and store the data. Through a process known as binning, the outliers are eliminated and the missing values are filled using the highest count of the encoded numerical values of a specified parameter. By eliminating duplicate values, this feature aids in maintaining data consistency. Training and testing data are two more categories into which the data set is separated. After the training data is entered into the model, it is trained. Values for the testing dataset may then be predicted by the model.
[0003] In some embodiments, inconsistent, noisy, and incomplete data are eliminated from the raw dataset through the data preprocessing procedure. Low-quality values are included in the raw data, which could have an impact on the data mining procedure. As a result, data preprocessing is crucial to the data mining procedure. Data preprocessing techniques include data reduction, data transformation, and data cleaning. However, years of education also had a significant impact on the anxiety of graduating undergraduate students. It should be mentioned that education is the process of educating or applying knowledge in the mind; in other words, it is a tool to teach the student the aptitudes, skills, and abilities required to function effectively in society. K-Closest neighbor (KNN), an administered arrangement strategy, is employed. Because it can function optimally for a variable number of boundaries in a short amount of time, this strategy is more qualified. Initially, a variable K is assigned and an irregular worth is anticipated. The Euclidean distance formula is used to calculate the smallest possible distance between the new record and each of the previous records. The principal K records are evaluated after all of the shortest distances are arranged in ascending order, and the new record with the highest all-out factor count is chosen.
[0004] In some embodiments, they retrieve the stress factors that are used to measure stress levels. Gender, age, financial issues, family issues, work schedules, learning methods, medical issues, favoritism issues, pressure issues, routine issues, communication issues, and so on. financial status, monthly income, and family economy, challenges teachers face in adapting to school, Years of education, physical activity, and self-evaluation of sleep quality. However, in stress studies, the following variables have been used most frequently: Age, physical, emotional, and depressive symptoms, feeling extremely worn out or exhausted, Parental occupation, financial status, monthly income, family economy, relationship with family, challenging environment, studio environment, immersive virtual environment, challenges with school adaptation, teachers, and self-evaluation of sleep quality. The information was collected, assessed, and prepared. The rules that have been put together include things like sexual orientation, financial difficulties, family issues, health issues, prejudice fix, stress, standard, and collaboration. These parameters are transformed into numerical values, which are then used to train the model. The "KNN classification Algorithm," a supervised learning machine learning technique, is used to predict stress. It is well known that this algorithm yields accurate results.
[0005] FIG. 2 illustrates a method for database search criteria according to an embodiment herein. In some embodiments, SVM is a machine learning technique that is specifically applied to problems involving regression and classification. Both continuous and categorical data can be handled by it. Data points that are represented on a hyperplane can be divided into two groups using SVM. A point belongs to the same group if its properties are similar. An essential step is the variable selection algorithm, which helps identify the variables with the highest correlation with the class and improves the accuracy of the methods employed. Variable selection, then, is the process of selecting which of a large number of variables to include in a specific model; in other words, picking suitable variables from a full list of variables by eliminating those that are superfluous or irrelevant. This module was used to import the testing dataset from a dominate sheet. The percentage of pressure in each testing dataset record is categorized as either peaceful or unpleasant, and then further described into the degree of pressure, which is shown with groups as previously illustrated. Additionally, the pressure level is visualized externally.
[0006] In some embodiments, one technique used in machine learning from the insights field is logistic regression. The technique aids in the resolution of classification issues. In essence, logistic regression and linear regression are the same process used to try to predict the output variable Y from the given set of X inputs. These are supervised learning techniques that use labeled data to train a large number of perceptions of both independent (X) and dependent (Y) variables in an effort to predict the responses of unlabeled, unseen data. It is a process for modeling and analyzing the relationship between variables and, frequently, how they work together to produce a particular outcome. It should be mentioned that the accuracy of methods for predicting stress and anxiety is correlated with the volume of data used in the studies. Support Vector Machine (SVM) and Multiple Layer Perceptron (MLP) were the algorithms that were able to predict anxiety with higher accuracy. At the same time, K-Nearest Neighbors (KNN) and Logistic Regression (LR) were the algorithms that produced higher accuracy in the stress case. It should be mentioned that this has to do with how much data was considered in the investigations. Students are shown and made aware of various boundaries. Understudies can enter these variables to gauge their level of anxiety. The gained result is separated into two classes: distressing and calm, and the unpleasant outcome is additionally partitioned into the level of pressure and outwardly showed.
[0007] In some embodiments, the Xg Boost is also known as the ensemble method or gradient boosting algorithm. It provides better performance for determining a range. It is a technique that aids in predicting the result by reducing residual errors in models. It is very helpful in preventing overfitting of the model, which helps to minimize loss. The most widely used method is Support Vector Machine (SVM), primarily because of its consistent performance in medical datasets where the ideal hyperparameters were identified and the effect of feature selection was assessed when building the model with a smaller feature set. In a similar vein, some studies have received higher reception for the percentage of anxiety accuracy, with Support Vector Machine and Multiple Layer Perceptron (MLP) showing the best results, with ranges of 0.6970 to 0.100 and 0.6750 to 0.9950, respectively. About 90% of the original dataset is made up of training data, and 10% is made up of testing data. By contrasting the expected results of the testing dataset with the original dataset, KNN calculates accuracy. The accuracy of the model increases with the number of records that are correctly predicted following matching.


, Claims:1. A method for machine learning-based approaches for predicting student stress in higher education, wherein the method comprises;
collecting relevant data from students, such as academic performance, attendance, engagement, personal and socio-economic information, is essential to predict stress levels accurately;
selecting and engineering the right features directly impacts the accuracy of the stress prediction model;
training on historical student data can effectively predict student stress levels;
evaluating metrics are necessary to ensure the predictive reliability and accuracy of the model;
integrating the model into a dashboard or app that provides counselors and support staff with real-time insights into student stress levels, allowing for timely intervention;
monitoring the model's predictions against actual stress outcomes; and
retraining the model with fresh data each semester or year to adapt to evolving student demographics and stressors.

Documents