ABSTRACT:
The study formulated a model for the determination of risk of depression among university students with a view to evolving a tool to assist mental health workers to help them make more consistent diagnosis of depression as well as identify students who need special attention. The algorithms underlying evolutionary computation (genetic algorithm and genetic programming) were used to formulate the model. Identification and formulation of the most relevant risk factors for depression was done using genetic algorithm (GA). The formulated model was simulated using R programming within the R Studio development environment. The performance of the model was validated using accuracy, true positive rate, precision and false positive rate.
The results of the predictive model developed using genetic programming for the two-class dataset without feature selection showed that 375 out of the 383 actual low risk and 90 out of the 124 actual high risk were correctly classified giving a total of 465 correct classification out of 507 with an accuracy of 91.7% while for the five class dataset without feature selection showed that 373 Normal cases out of 383, 17 out of 56 low risk cases, 0 out of 25 medium risk cases, 22 out of 27 high risk and 10 out of 16 very high cases were correctly classified giving a total of 422 correct classification out of 507 with an accuracy of 83.23%. Also, for the predictive model developed using genetic programming and feature selection, the following results were observed. For the two-class dataset, 380 out of the 383 actual low risk cases and 95 out of the 124 actual high risk cases were correctly classified giving a total of 475 correct classification out of 507 with an accuracy of 93.7% while for the five-class dataset, 373 Normal cases out of 383,17 out of 56 low risk cases, 0 out of 25 medium risk cases, 22 out of 27 high risk and 10 out of 16 very high cases were correctly classified giving a total of 422 correct classification out of 507 with an accuracy of 83.23%.
The study concluded that the resulting model can be integrated into existing Health Information System (HIS) which captures, stores and manages clinical and demographic information about students. Such information can be fed to the depression risk prediction model, thereby providing aid to physicians and improving clinical decisions affecting depression.
Keywords:
Depression, Genetic algorithm(GA), Genetic programming(GP) and Predictive model
