Abstract: The prediction accuracy of most current software reliability prediction models is not high. This paper put forward a software reliability prediction model based on AGA-LVQ,which takes advantage of non-linear computing power of the learning vector quantization (LVQ) neural network and parameter optimization capability of the adaptive genetic algorithm (AGA). Firstly,principle components analysis (PCA) preprocessing was used to reduce the dimension of the metrics and remove the redundancy and error data. Secondly, AGA was used to calculate the optimal initial vector weights of the LVQ neural network. Lastly, LVQ neural network was used to do the software reliability prediction experiments. The experiment results indicate that the method has a higher prediction precision than the traditional software reliability prediction model.

Key words: Software reliability prediction, Pattern recognition, LVQ neural network, AGA, PCA