Systematic review of machine learning techniques to predict anxiety and stress in college students

Background: Anxiety is considered one of the most common pathologies that people go through frequently, this being the main cause of illness and disability in students since it is more common in women with 7.7% than in men with 3.6%. Moreover, stress is also one of the main causes of some health-related problems, such as cardiovascular diseases and mental disorders. Objective: The purpose of this study is to gain a deeper understanding of the methodologies, attributes, selection algorithms, as well as techniques, tools or programming languages, and metrics of machine learning algorithms that have been applied in the prediction of anxiety and stress in college students. Methods: An exhaustive search of 29 articles was performed, using keywords from 7 databases: ScienceDirect, IEEE Xplore, ACM, Scopus, Springer Link, InderScience and Wiley from 2019 to 2023. This article was based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) methodology, taking into account the inclusion and exclusion criteria. To then make a synthesis of the findings of the studies about the following aspects such as methodology, attributes, selection algorithms, as well as techniques, tools or programming languages and metrics. Results: The methodology most used was based on the sequence of steps, the most important attributes were age and gender, also most studies do not use variable selection techniques; on the other hand, the most efficient techniques were Support Vector Machine (SVM) and Logistic regression (LR), the most used programming language to develop the models was Python and finally the essential metrics to determine the effectiveness of the model were Precision and Accuracy. Conclusions: This systematic review provides scientific evidence, with results describing how machine learning techniques help predict anxiety and stress. For this, machine learning algorithms are compared to perform a broad analysis of these algorithms, tools or Programming languages, metrics, selection variables and influential factors, which will help in medical fields for detection of anxiety and stress.