Depression Prediction Using the Classification and Regression Tree (CART)
Keywords:Depression, data mining, training and testing, classification, regression tree
Depression is a mood disorder that involves the continuing feeling of sadness and loss of interest. The crucial life events for an individual, such as losing a job may lead to depression. However, the feelings of grief and sadness are clinically diagnosed as part of depression only if the symptoms persist for at least two weeks. Eventually, depression can last for several weeks, months, or years. Some symptoms of depression may overlap with other somatic illnesses and cause difficulty in diagnosing it. This research aims to use the developed forecast model to predict future depression cases and it uses classification and regression tree (CART) of data mining approach, to predict or classify whether an individual suffers from depression or not. The dataset that was used in this research is the depression dataset from the Dataset of Students’ Mental Health at an international university in Japan. This dataset consists of 268 numbers of instances and it has 10 attributes. In addition, to acquire the results, the machine learning software that was used is R Studio and the language that was used is R Programming. Besides that, evaluation metrics were used to evaluate the performance of the forecasted model and the evaluation metrics that were used were accuracy, precision and recall. From the research, it shows that the value for accuracy is 0.50(50%), precision is 1.00 (100%) and recall is 0.50 (50%). Following that, it shows that this forecasting model has the highest value of precision which is 1.00(100%). Furthermore, from the data, it also shows that teenagers in the age range from 18-22 are most likely to get depression and they also have the intention of suicide. Lastly, in the future, this research could be continued with more training on different datasets and more different techniques could be used. Besides that, this research could be improved by adding other algorithms to best understand the strengths and weaknesses of other techniques.