Modeling software failures using neural network

Failure-free software is a major concern for delivering high-quality system. High reliable software system requires robust modeling techniques to estimate the probability of the software failures over a period of time. In this paper, we have proposed a neural network based approach for predicting software failures. This paper presents the use of Feedforward neural network, mostly adopted by many researchers for reliability prediction [12][13][14] and Elman neural network. This experiment was conducted on real software failure dataset for three different applications. We have used various error metrics such as MSE, RMSE, NRMSE, MAE, MAPE, MMRE and BRE for the performance analysis of Feedforward (Universally adopted) and Elman neural network. Experimental results show that Elman neural network has good predictive capability. Mean Square Error (MSE), Mean absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) are both reduced significantly for Elman Neural Network in comparison to Feedforward network.