Software Evolution Prediction Using Fuzzy Analysis

摘要

Software evolution prediction has been one of the objectives in the past studies on open source software. A perusal of the existing research in this area shows ARIMA modeling as the most frequently used prediction procedure. This paper analyses the ability of a method of prediction based on fuzzy time series to predict the next state of data. The data set is related to commit history of three open source software projects as recorded in the revision control tool Git. The prediction method outperforms the ARIMA (0,1,0) -- the simplest benchmark model.