Design and implementation of a personal loan default prediction platform based on LightGBM model

The LightGBM algorithm is used to provide new research ideas for the analysis of customer data information in the practical application of personal loan default models, in view of the fact that financial institutions are not able to reasonably assess the default risk in each lending. By comparing the accuracy, precision, recall and ROC curves, we found that the personal loan default model constructed by the LightGBM algorithm has better prediction results than the personal loan default prediction model constructed by the Random Forest algorithm, with the prediction accuracy reaching Then, based on the personal loan default model constructed by the LightGBM algorithm, a personal loan default system platform was built using the python language and the Flask framework, which was able to achieve real-time input of customer information data to predict the probability of whether the customer would default or not.

Перевод пока недоступен