Empirical Analysis on Mental Health Prediction Using Image Processing and Machine Learning Approach

Abstract

A vital component of wellbeing is mental health. Our mental health is very important as it maintains our overall well-being, affects how we think feel and act, and prompt treatments can be made when mental health issues are detected early.Empirical analysis ensures that findings are supported by actual evidence by employing measured and observed data to make inferences. This approach enhances the accuracy and reliability of predictions in data-driven studies. The goal of this research is to support mental health diagnostics by predicting mental health disorders through image processing techniques. Using a robust dataset having four features such as Talking, Walking, Eating, Playing, a variety of machine learning models were applied in the empirical investigation to reliably categorize mental health problems. The models' performance was assessed using metrics for accuracy, recall, and precision. A variety of techniques are employed, including ensemble approach, KNN, Activation Function, ANN, LSTM, CNN, Decision Tree, Random Forest, SVM, and Logistic Regression. This research has obtained accuracy about 0.995 for logistic regression, 0.955 for decision tree, 0.994 for random forest, 0.959 for SVM, 0.993 for ensemble methods, 0.988 for KNN, 0.999 for activation function, 0.989 for ANN, 0.988 for LSTM, 0.988 for CNN regarding talking feature. This research has obtained accuracy about 0.979 for logistic regression, 0.988 for decision tree, 0.987 for random forest, 0.98 for SVM, 0.992 for ensemble methods, 0.992 for KNN, 0.994 for activation function, 0.931 for ANN, 0.994 for LSTM, 0.956 for CNN regarding walking feature. This research has obtained accuracy about 0.979 for logistic regression, 0.981 for decision tree, 0.999 for random forest, 0.922 for SVM, 0.963 for ensemble methods, 0.941 for KNN, 0.987 for activation function, 0.995 for ANN, 0.91 for LSTM, 0.982 for CNN regarding eating feature. This research has obtained accuracy about 0.98 for logistic regression, 0.975 for decision tree, 0.994 for random forest, 0.98 for SVM, 0.91 for ensemble methods, 0.978 for KNN, 0.967 for activation function, 0.972 for ANN, 0.993 for LSTM, 0.996 for CNN regarding playing feature.