Software Recession Prediction Algorithm

Measurement-based software rejuvenation strategy is a process that includes monitoring and collecting parameters of system performance periodically, forecasting and evaluating the speed, time and possibility of software aging and the cost of rejuvenation so as to determine whether to carry out software rejuvenation or not and when.This paper discusses measurement-based software rejuvenation strategy, combined with wavelet analysis and neutral network methods to analyze the parameters of system performance which are monitored and colleted periodically. In conclusion, it predicts resource consumption and the trend of software aging.This paper makes an improvement of threshold de-noising method through data de-noising functions based on wavelet analysis. Gradient-based self-adaptive learning algorithm is used to select threshold. The experiments show that this method has better performance of de-nosing than others. It can remove noises and also guarantee the precision of signal at the same time.As to the forecast of preprocessed data, this article first adopts wavelet decomposing algorithm to decompose data to each scale coefficient and then constructs a BP neural network to predict each scale coefficients. Finally, the coefficients are reconstructed to predict the results. LM learning algorithm is adopted in BP network learning. A new stimulation function is improved for reducing learning time and an improved method is also introduced to determine neuron numbers of hidden layer. Meanwhile, focusing on all the wavelet coefficients of data, another direct forecast method is proposed. The experiments show that both methods are fast and efficient. In addition, the forecast results using these methods can meet our demands.