An Improved Bagging Ensemble in Predicting Mental Disorder using Hybridized Random Forest - Artificial Neural Network Model
Oluwashola David Adeniji, Samuel Oladele Adeyemi, Sunday Adeola Ajagbe
Abstract
Machine Learning majorly provides the process of collecting, identifying, pre-processing, training, validating and visualization of data. This study identifies the problem of late detection of mental disorders in IT employees. There are many cases of mental disorders that are not apparent, notable or diagnosed until they become critical. This affects the productivity of the employees not only in IT industry. The objective of the study is to develop a Hybrid Random Forest (RF) and Artificial Neural Network (ANN) model to predict mental health disorders among employees in an IT industry. The experiment applied a hybrid Random Forest and Artificial Neural Network (RF-ANN) model in predicting the chances of IT employees developing mental disorders. To measure the performance of the model, RF and ANN algorithms were separately developed, their results were recorded and compared with the results of the hybrid model. In the hybrid model using “Bagging Ensemble,” the prediction of an IT employees developing Mental Disorder shows the weighted average performance of 84.5% for precision, recall, and accuracy and precision is 82.5% using the hybridized RF and ANN models on “Bagging Ensemble”. This result obtained from the hybrid model correctly shows a significant improvement in its performance over individual performances of the RF model and ANN models. There was a marginal improvement in the performance of the hybrid model when compared with the result of the parameter-tuned RF. This suggests that by applying the RF-ANN model an improved dataset could be investigated and compared with results obtained in this study.