Contributing to diagnoses of Mental Disease Using New Optimization Machine Learning Methods

Abstract

Mental illness is known to be difficult to diagnose and we can say that if someone has a mental
illness, it can affect someone for years before diagnosis. Geriatricians often encounter a large number of
patients for treatment without being able to reduce or automatically diagnose them. Designing a system to help
better to diagnose this disease and to reduce cost and time is our aim in this study. We used a mental dataset
applied with data mining optimization algorithms and we applied it with the Python programing language,
including training the test split and pre-processing feature selection model used random forest (RFE) and
feature importance to enhance the system results and accuracy for mental dataset in our research and processed
the missing values that found with attributes.The best accuracy was achieved by Adaboost optimization model,
which gave us 99% accuracy and the Adaboost ensemble merged with the decision tree produced a 94%
accuracy. Moreover, the random forest optimization produced better accuracy at 96% and 92% resulted from
using the SVM algorithm. Finally, this optimization system by merging the two algorithms to work together
will be more efficient and better able to help classify and diagnose suffering patients using a huge amount of
data in little time and at low cost.