Implementation of an Artificial Neural Network Algorithm for Mental Illness Virtual Assistant Chatbot Development

Abstract

Mental health is a critical issue in modern society, yet access to psychological support remains limited. This study presents the development of a chatbot as a virtual assistant for individuals experiencing mental illness using the Artificial Neural Network (ANN) algorithm. The dataset was manually constructed and divided using an 80:20 ratio for training and testing. The ANN model employs one hidden layer with ReLU and softmax activation functions to classify user input into relevant mental health categories. The model achieved a training accuracy of 83.2% with a loss of 0.655, and a testing accuracy of 81.5%, indicating solid performance. Compared to rule-based methods, ANN provides better adaptability in recognizing diverse expressions and delivering context-aware, empathetic responses. This study also introduces a custom-built mental health dataset and integrates a crisis response module that is underexplored in previous research. The chatbot targets five categories of mental disorders: Schizophrenia, Bipolar Disorder, Depression, Anxiety, and Personality Disorders. Findings suggest that ANN-based chatbots can serve as reliable, accessible, and scalable early-stage mental health support tools.