An Optimized Machine Learing Framework For Extracting Suicide Factors Using K-Means++ Clustering

Suicide has emerged as one of the serious problems which should be eradicated from the society. People with suicidal thoughts restrict themselves by not expressing thoughts to the people around them. Studies have shown that people show more interest in expressing their thoughts over social media platforms. So, research has been conducted to identify people with suicidal ideation by analyzing the posts which they posted in social media platforms. Certain studies mined out new factors which influenced people to commit suicide, but those factors had certain drawbacks in it. This paper mainly focuses on overcoming those drawbacks in the factors. A new modified approach for extracting those risk factors is introduced as it can be used for future works related to suicidal ideation detection tasks. Statistical methods were imposed on the data to mine out the underlying characteristics of the features. K-Means++ clustering algorithm was implemented to extract the modified features. The modified features were given as an input for a testing classifier, and it attained an accuracy of 75.13%.