A Hybridized Feature Extraction Approach To Suicidal Ideation Detection From Social Media Post

Suicide's been a rising social problem. Concerned society has expressed worries regarding the recent increase in committing suicide. There are different stages of the suicidal act. If one can get recovery from early-stage which is, suicidal ideation, it is possible to reduce the number of suicides per year. Our study aims to detect suicidal ideation from social media-based using Natural Language Processing (NLP). We have applied the best feature extraction methods- Genetic and Linear Forward Selection (LFS) to select the best features from our feature vectors using our proposed formula. Moreover, We have created a robust feature set based on different computational and linguistic features. Finally, we have shown that by applying our hybrid feature extraction method we can get a significant increase in our accuracy to detect ideation.