Support Vector Machine Parameters Optimization using Firefly Algorithm for Mental Health Prediction

The increase in mental health problems has prompted Machine Learning to approach efforts to accurately diagnose mental health. One technique that is frequently used in the classification process for signs of mental health issues is the Support Vector Machine (SVM) method. The SVM technique relies on being able to ascertain a kernel function’s parameter σ and soft margin coefficient C. Usually these two parameters are obtained using a fairly long method, namely a trial-and-error mechanism trying to combine every possible SVM parameter. The aim of this research is to determine the most optimal value of an SVM parameter based on the level of accuracy and time by combining SVM with the Firefly Algorithm (FA) optimization method. Response data on employee mental health at various technology companies was used for case studies. There is a division of 1400 data into 200 data as test data and 1200 data as training data. In the evaluation stage, the Firefly Algorithm - Support Vector Machine (FA-SVM) method has the ability to achieve optimal parameter combinations with higher accuracy than the trial-and-error method. With $C=1.0-3.0$ and $\sigma=\mathbf{0 . 1 - 1 . 0}$ as the value ranges, the SVM evaluation yields 77% accuracy. The FA-SVM evaluation results, on the other hand, indicate the highest accuracy of 87.84%. The FA-SVM takes only 2330 seconds to execute, compared to the SVM method’s 5928 seconds; this represents a 3598-second difference in execution time.