Machine Learning Model for Heart Failure Prediction with Feature Selection and Data Expansion

Heart failure is a serious cardiovascular disease whose treatment and management require accurate predictive models. However, traditional feature selection methods may face difficulties in dealing with heart failure data, especially when the data size is small. Therefore, this thesis combines the large language model to overcome the drawbacks of small sample data, while this study proposes a new collaborative filtering-based feature selection method, which is intended to be used in the construction of heart failure machine learning models to alleviate the preference of different models for feature selection in order to ensure that the selected features have better generality. In the study of experimental data from 299 patients, we used visualisation techniques to improve the transparency and credibility of the scheme, and the experimental results further validate the feasibility and efficiency of the scheme. This research result provides new support and guidance for the prediction and treatment of heart failure.