A Novel Shuffle Encode Distil Model for Suicide Risk Detection in Social Media

Depression leading to suicidal tendency has become an important issue worldwide. However, a large number of people generally, lose their life undiagnosed and untreated. Here a social media platform with machine learning provides the vital solution to this issue. Many studies have shown that men with depressive disorder tend to post depressive messages on these platforms. This paper proposes a machine learning Shuffle Encode Distil model which can identify and classify the messages and posts as suicidal and non-suicidal. The proposed model was also tested against a few state-of-the art machine learning models like XGBoost, Random Forest, CatBoost, Support Vector Machine Classifier, etc and other proposed models. The Shuffle-Encode Distil-Model was able to classify the messages successfully with accuracy of .999, precision of 0.998, recall of 0.999 and F1 score of 0.999.