Abstract:

Non-suicidal self-injury (NSSI) is a significant public health problem characterised by widespread stigma, high complexity and heterogeneity. Traditional NSSI research measure and analysis methods are limited, resulting in very low predictive power of the identified factors. In recent years, machine learning has gradually been applied to the analysis and modelling of NSSI. Through simplified questionnaire models and complex multimodal data models, the importance of predictive factors can be visualised and more accurate NSSI classification can be achieved, thus improving the overall predictive performance to a moderate level. In the future, it is necessary to combine traditional NSSI theories and methods to make the screening criteria more stringent, and combine unsupervised learning with transfer learning to increase the reproducibility and comparability of the models. Furthermore, combining non-questionnaire NSSI data with deep learning meanwhile is helpful to improve the predictive performance.

Key words: machine learning, non-suicidal self-injury, predictive power, application