Techniques for Suicidal Ideation Prediction: a Qualitative Systematic Review

Suicide is an increasingly present issue in our society whose eradication could be greatly aided by decision support technologies that can objectively identify early markers of suicidal ideation. We present our paper which reviews various existing techniques for suicidal ideation prediction. These techniques are broadly divided into two major categories – 1) Text-based indicators (which employ NLP) and 2) Behavioural indicators (like eye gaze and smile which are classified using ML algorithms). The existing techniques aim at classifying individuals as suicidal or non-suicidal using classifiers such as Random Forest, Logistic Regression, Naïve Bayes, K-Nearest Neighbour (KNN), Support Vector Machine (SVM), Multi-Layer Perceptron and XGBoost. Each technique has its advantages and disadvantages. We aim to bring all these diverse techniques to one place and try to show how each one of them is individually contributing to prediction. We also shed light on how combining these techniques and eliminating the disadvantages could lead to better suicide prediction.