Depression status identification using autoencoder neural network.

Process of Depression Detection Using Auto Encoder Network. [Display omitted] • An efficient machine learning algorithm for depression status detection based upon EDA is proposed. • The model is based on emotional context insensitivity during depression. • A set of features from different signal representations is extracted. • Optimal features subset and network configuration is identified. Depression is the leading mental illness/disorder around the world with global number reaching up to 300 million worldwide. This mental disorder is more prevalent in youngster of age group of 18–25 years especially in developing countries. Early diagnosis and treatment are required to this alarming problem specifically for an adolescent student suffering from these disorders who more often goes undiagnosed and hamper their progress at the critical movement of their life. The cheaper automatous system such as electrodermal (EDA) can help in with early diagnosis of depression disorder. In this paper, EDA based machine learning using autoencoder network (AEN) and deep neural networks (DNN) was developed for detecting level of depression among 38 university students. Developed AEN and DNN algorithm was able to classify five categories of depression with training of 96.5% and 94.5%, testing accuracy 95.2% and 94.2% while overall network accuracy was 94.0% and 92.0% with high sensitivity and specificity rate. [ABSTRACT FROM AUTHOR]