Back

A Machine Learning Perspective on Causes of Suicides and identification of Vulnerable Categories using Multiple Algorithms

Shreemali, J.; Chakrabarti, P.; Chakrabarti, T.; Poddar, S.; Sipple, D.; Kateb, B.; Nami, M.

2021-04-13 psychiatry and clinical psychology
10.1101/2021.04.08.21255162 medRxiv
Show abstract

BackgroundSuicides represent a social tragedy with long term impact for the family. Given the growing incidence of suicides, a better understanding of factors causing it and addressing them has emerged as a social imperative. Material and MethodsThis study analyzed suicide data for three decades (1987-2016) and was carried out in two phases. Machine Learning Models run after pre-processing the suicide data included Neural network, Regression, Random Forest, XG Boost Tree, CHAID, Generalized Linear, Random Trees, Tree-AS and Auto Numeric Model. Results and ConclusionAnalysis of findings suggested that the key predictors for suicide are Age, Gender, and Country. In the second phase, data from happiness reports were merged with suicide data to check if Country-specific factors impact the list or order of key predictors. While the key predictors remain the same, Country-specific factors like Generosity, Health and Trust impact the suicide rate in the Country.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.