Prediction of 30-Day Readmission: An Improved Gradient Boosting Decision Tree Approach

The rate of 30-day hospital readmission is commonly used to measure the quality of medical service. To improve the hospital efficiency, in this study, we have proposed a balanced fuzzy C-Means LightGBM classifier (BFCMLGB) to predict the 30-day patient readmission using a published readmission dataset. The results show that the BFCM-LGB algorithm can effectively process the problem of the imbalanced readmission data, and further conduct a feature extraction and a parameter optimization. Compared with other known algorithms, BFCM-LGB performs the best, in terms of four measurements, accuracy, G-mean, F-score, and area under the curve. Altogether, BFCM-LGB can be a useful classification algorithm to predict 30-day patient readmission.