An Empirical Systematic Study of Deep Learning-Based Software Defect Prediction Models in Complex Systems

This paper introduces an empirical study on a complex system environment, aiming to construct and evaluate a deep learning-based software defect prediction model. The model innovatively integrates Long Short-Term Memory (LSTM) network, which uses LSTM's powerful time series analysis ability to mine potential defect patterns for historical data accumulated during software development. The model can capture the trend of software defects over time and cover multiple dimensions. While maintaining the integrity of time series information, the model enhances the expression ability of high-dimensional feature space, which provides a new technical approach for defect prevention in complex system environment.