Software Usage Prediction Based on Hybrid LSTM-ARIMA Algorithm
Yuan Wang, Ying Nan Zhou, Keyu Yan, Yujie Liu, Ruixiang Huang, Tianyi Wu · 2023
The number of users of software is an important indicator for engineers when developing, modifying, and maintaining software. However, the number of software users is often influenced by numerous factors, such as software features, trends, and so on. This leads to the fact that the number of users contains both linear and non-linear trends in the process of change. This leads to the original single algorithm not being able to predict the number of users of the software better. For this reason, a hybrid LSTM-ARIMA algorithm is designed in this paper to predict this problem. In this paper, a Bayesian combination model is used to calculate the weights of two single algorithms respectively, so as to realize the combination of two algorithms. Experiments show that the designed algorithm can better fit the trend of the change in the number of people and has a low SMAPE, which indicates that the prediction performance of the algorithm is good.