Психикалық денсаулық көрсеткіштерін талдау: депрессия болжамын интерпретациялау үшін машиналық оқытуды қолдану

Abstract

Abstract. This research paper explores the theoretical and applied importance of using artificial intelligence methods to analyze mental health indicators. The main objective of the study is to conduct a comparative evaluation of the effectiveness and interpretation of several machine learning models in order to improve the overall quality of clinical decision-making. To achieve this, widely used machine learning algorithms, including logistic regression, random forest, gradient boosting methods, and neural network models, have been selected. The SHAP method has been used to determine how models generate predictions and to identify the most influential variables. This approach allows for the quantification of each feature's contribution to the model's final outcome and provides a mechanism for interpreting the internal logic of algorithmic assumptions. By decomposing the model's results into individual descriptive effects, SHAP facilitates an open understanding of complex predictive systems. Empirical evidence suggests that models with more complex architectures, particularly neural networks, achieve relatively high levels of prediction accuracy. In addition, these models exhibit limited interpretability, making it difficult to understand their prediction logic. In other words, while these algorithms may produce highly accurate results, their decision-making process often remains unclear to clinical professionals. In contrast, ensemble algorithms such as random forests and gradient boosting have shown a favorable balance between predictive performance and model transparency. These methods have provided reliable results when applied to clinical data sets, allowing for meaningful explanations of the impact of various variables on predictions. The analysis also shows that integrating interpretable machine learning methods into medical data analysis can increase the trust of psychiatrists and healthcare professionals in AI-based decision support tools. Therefore, improving the transparency of predictive models is an important step towards the responsible and effective implementation of AI technologies in clinical practice.