DCNN-WiGRF: Depression Recognition through Deep Feature Extraction using Deep Learning and weight updated information gain Random Forest technique

Abstract

Depression is a general illness globally which causes severe implications. Depressive symptoms can predict earlier and it is considered as initial step in evaluation and treatment. There is a potential to generate computational intelligence model to detect the depression symptoms by using the depression datasets and machine learning advancements. Therefore, this research proposed DCNN-WiGRF algorithm to identify the depression disorder types such as normal, persistent disorder, bipolar disorder, and MDD. Initially the data is pre-processed and then augmented by using data augmentation process in which the rotation of image is performed. Further, the features are extracted by using Deep CNN algorithm from FER2013 dataset. Finally, the classification of the significant features is performed by using the Weight updated Informative Gain Random Forest algorithm based on queries and extracted features. Here the information gain is used because if the faster ability and it can exactly predict the depression from non-linear data. Thus, the depression prediction model can be adopted, which can contribute to classify the depression disorder types in order to provide earlier treatment.