Research and Application of Project Settlement Overdue Prediction Based on XGBOOST Intelligent Algorithm

In order to strengthen the management of the company's project settlement, as well as ensures the completion of project settlement on time and quality according to the completion of the project, to improve the project management level, and strengthen the responsibilities, this paper analyzes the reasons for the project settlement overdue and proposes a project settlement based on XGBOOST algorithm. Overdue prediction model. Based on the historical engineering registration form and ERP data, the model sample data set is obtained through data predicting technology, and the XGBOOST prediction model is established. According to the relevant information of the project, the number of days in the settlement of the new completion project is predicted, and the standard days specified in the project management can be compared to the number of days exceeded, so that the relevant person in charge of the project overdue can be issued an early warning for 17 to 50 days in advance. The actual test results show that: The predicted days of the completion of the delivery of materials, non-material settlement, material settlement and asset creation are basically in line with the actual value, and the average absolute error rates of the predicted values are 14.55%, 8%, 11.38% and 8.25%, respectively.