Early Detection of Suicidal Predilection on Instagram Social Networking Service

Social Networks has become the biggest medium of expression of one’s thoughts and emotions. With more than one-third of the global population expressing their thoughts, opinions and events, these networks become rich with direct indicators about the subject. Multimedia based social networking services are gaining popularity than text-based ones. In this work, we analyze the data from Instagram, which is one of the largest image and video-based social networking service with more than one billion users. This paper analyzes the social media posts on Instagram with the goal of finding when the subject feels low that could potentially lead to a suicidal tendency in the initial stages. In this study, the Instagram posts were analysed using different features that are more inclined to exhibit suicidal behaviour. The best accuracy is achieved by J48 binary tree classifier with a classification accuracy of 87.68%.