Abstract:
Problem: There are various reasons that often lead to the neglect of mental health issues. These
include the aspects of social stigma, lack of time, and limited scope of traditional face-to-face
counseling. Most of the online counseling sites available do not utilize machine learning to
determine the probable risk of mental health issues arising from certain user’s symptoms nor do
they give recommendations of suitable counselors as well as put in place strict confidentiality. This
project seeks to overcome such deficiencies through the creation of an online counseling platform
which is able to anticipate mental health disorders from the users’ symptoms and suggest suitable
counselors.
Methodology: An Agile development approach quickly validated the design through the use of
both quantitative and qualitative methods. A machine learning structure is embedded in the
platform, which will train on a symptom-disease database, and this enhances the predictive ability
of the platform. User surveys and interviews were conducted to elicit requirements so that the
platform will effectively address user needs regarding ease of access and confidentiality. Areas of
concern during development involved feature development and training and testing of predictive
models to enhance precision in the predictions and user interest.
Initial Results: The results obtained by the prototype were encouraging, with a prediction
accuracy of mental health conditions at 89%, an AUC-ROC of 0.91 and a 5% false positive rate.
These outcomes corroborate the platform’s reliability in offering predictive matrices for mental
health disability and recommending counselors’ services to users at first informs them about the
mental health problem aiding in identification. Further testing is ongoing to enhance model
precision and lower the occurrence of false positive rates.
