Suicide Ideation Detection Using ML and DL Algorithms Assisted by NLP Techniques

This manuscript investigates the efficacy of machine. learning (ML) and deep learning (DL) algorithms, augmented by Natural Language Processing (NLP) methodologies in identifying instances of suicide ideation within textual datasets. In the context of an escalating global mental health crisis, the precocious detection of suicidal ideation is paramount. Utilizing extensive datasets harvested from social media and various digital forums, this research endeavors to forge predictive models capable of discerning indications of suicidal contemplations. Such models aim to furnish mental health practitioners with the means for opportune interventions. Comparative analysis within this study reveals that DL-based algorithms, particularly when employing Word2Vec for token vectorization, outstrip conventional ML approaches in accuracy. Conversely, the Näıve Bayes algorithm exhibits superior performance with TF-IDF vectorized data. Furthermore, an ensemble approach to ML algorithms marginally enhances performance relative to singular algorithm applications.