Role of Machine Learning in Short-Listing Future Suicidal Candidates


  • Summyia Sadia, Fariha Tariq, Pervaiz Zarif, Kishwar Naheed, Abid Rafiq, Hafiz Muhammad Faisal



Machine Learning, Suicide Prevention, Suicide Detection and Ideation.


Suicide is an important issue to address, especially in rural areas. Rural areas are facing unique challenges such as poor health care facilities, lack of awareness, financial constraints and many more for such matters.

Aims: To find the social, educational and medical attributes which may lead a person to deliberate self harm.

Study Design: Retrospective study.

Methodology:  Total 100 cases of suicidal attempts taken from DHQ teaching hospital Sargodha from (June to December) 2019. We considered all the suicidal and self harm cases admitted through emergency and medicolegal clinic. Moreover cases less than 9 years of age and autopsy cases were excluded. All the cases were analysed with reference to 10 features (age, gender, locality, education, marital status, duration of stay in hospital, treatment given, prevalence of psychiatric disorder, suicidal attempts, the method used for suicidal attempt).

Statistical analysis: ML models work on numeric data. However the dataset we collected have categorical features except age. The most used method for such purpose is python get dummies function. The get dummies() function is used to convert categorical variable into dummy/indicator variables.

Results: In this study, more preponderance of suicidal attempts at age less than 40 in males which shows the development of more mature attitude with increasing age.

Conclusion:  It was concluded that suicide is influenced by many personal factors that cannot be shared publicly on social platforms. However, such information can be used to lower the risk of suicide attempts in rural areas.