A Review of Overfitting Solutions in Smart Depression Detection Models

Overfitting is a common issue in machine learning-based depression detection model. Overfitting occurs when a machine learning model uses garbage data in the training datasets to the amount that deleteriously impacts on the concert of the model on new data. To predict depression level, most of the model uses machine learning and deep learning approaches. In this paper, a survey is conducted on the solution of overfitting in machine learning-based depression detection models. The survey covers the latest research articles and thesis published in the years 2020 and 2021. The findings of the survey reveal that the early stopping methodology, also called Control the Iteration as is better as compared to the penalty methodology, as it can avoid overfitting problems in the machine learning-based depression detection model. We have listed down other methods also to reduce overfitting like regularization methods, applying ensemble learning, training with more data, cross-validation, etc.