Suicidal ideation prediction in twitter data using machine learning techniques

People prefer new technology by using online social media as a communication channels to express their suicidal thoughts. Primary identification and detection are viewed as an effective approach to avoid suicidal attempt and suicidal ideation- two basic hazards causing effective suicide. This paper exhibits different techniques to comprehend suicidal ideation through online user contents in particularly by considering twitter data for past last two years as an objective of early detection by means of sentiment analysis and supervised leaning methods. Analysing the text descriptions and users language exposes rich knowledge that can be utilized as a primary cautioning system for suicidal detection. To identify tweets exhibiting suicidal ideation, several features are extracted and a set of features are proposed for training the model over the dataset by using ensemble and baseline classifiers. Based on the outcome of baseline classifier; improved ensemble random forest (RF) algorithm achieved an accuracy of 0.99% compared to other classification methods for suicidal prediction with tweets containing suicidal thought is better when compared to the existing system. Such experimentation and monitoring may help individual and population-wide prevention by counseling and informing to suicidal research and policy. The experimental analysis expresses the feasibility of the methodology used by providing a benchmark for suicidal detection on online social network: Twitter.