Online Test System Load Balancing Method Based on Learning Analytics and Prediction Model

Abstract

Aiming at the problem of load imbalance such as database instability caused by high concurrency in the online test system, an online test system load balancing method based on learning analysis and prediction model is proposed. The method is divided into two steps. The first step is to build a curve of the number of questions answered according to the user test data. Simulate the test load change at different time intervals, reduce the number of concurrent connections by reasonably arranging the test; the second step is to analyze the user’s usual learning data and extract multiple features and then construct a time prediction model through the convolutional neural network. When the user accesses the test system, we can test the time required for him to complete the exam based on the predictive model. Each type of user is evenly redirected to each server by length of time. The experimental results show that the number of concurrent connections in the database generated by this method is lower, and the load of each server can be effectively balanced. In addition, the method can predict the load situation in advance so that can provide a reference for purchasing server resources before the test.