Combination of artificial neural network and fuzzy logic system: an algorithm for identifying, classifying, and comparing levels of divorce-related depression in older men and women

Depression due to divorce in the elderly is a negative psychological response. Diagnosis, classification, and comparison of divorce-related depression levels in older women and men have received less attention.

The aim of this study was to automate data processing to identify, classify and compare levels of depression due to divorce in older men and women.

The present study is an applied descriptive-survey research. The study was conducted in the elderly community seeking divorce in Gonbad Kavous. Before divorce, data were collected from the research community using the Geriatric Depression Scale. Fifteen normal and available couples with the necessary conditions were included in the study as a sample. Four months after the divorce, data were collected from sample couples using the Elderly Depression Scale and then randomly divided into training and experimental groups. MATLABR2016b software was used to implement artificial neural network with fuzzy logic approach. The ANFIS system was trained and then tested separately for older men and women. System performance was determined using the ROC curve. The prevailing level of depression after divorce in women was compared to elderly men.

Evaluation of ROC curve with TPR ˃94/66%, FPR ˃94/31% and AUC ˃0/84 (P= 0/000) showed high accuracy of the model. In addition, men were more likely to be depressed than women after divorce.

During the primary care program, in addition to using the system designed as a screening tool for early detection of psychological disorders in the elderly to increase the accuracy of specialist diagnosis, the possibility of appropriate and timely treatment for depressed men and women will be provided