T-Wave alternans (TWA) is a cardiac phenomenon regarded as an index of high risk of sudden cardiac death (SCD). Although a number of methods have been proposed for TWA detection, their final decision always fully depends on one single lead which is usually picked out for its strongest TWA detection result among all the other leads. That is to say that lots of useful information have been unused and wasted. To our best knowledge, no method that fuses TWA detection results independently obtained by each lead to do comprehensive decision has been introduced. In this paper, a novel multi-lead method for TWA fusion detection is proposed. This method combines Dezert-Smarandache theory (DSmT) with Laplacian likelihood ratio method (LLR) to gain higher detection rate. The proposed method was evaluated and compared with standard LLR method by means of a simulation study, in which extracted TWA waveform and clean background ECG from real records were used to synthetize simulated data with different types of simulated or physiological noises. These methods were also applied to real records from PTB Diagnostic ECG Database. The results are presented by constructing receiver-operator characteristic (ROC) curves. All test results show that the proposed method has a larger margin of separability and higher detection rate than traditional methods. A more accurate and robust TWA detection method has great value to predict the risk of sudden cardiac death(SCD).

In the software reliability growth phase, the nature of the failure data is, in a sense, determined by the software testing process. A hybrid model is proposed for medium and long-term software failure time forecasting in this paper. The hybrid model consists of two methods, Singular Spectrum Analysis (SSA) and Auto Regressive Integrated Moving Average(ARIMA). In this model, the time series of software failure time are firstly decomposed into several sub-series corresponding to some tendentious and oscillation (periodic or quasi-periodic) components and noise by using SSA and then each sub-series is predicted, respectively, through an appropriate ARIMA model, and lastly a correction procedure is conducted for the sum of the prediction results to ensure the superposed residual to be a pure random series. The software failure data of two real projects are analyzed as case studies. The results have been compared with the predictions made by ARIMA and Singular Spectrum Analysis-Linear Recurrent Formulae (SSA-LRF). It shows that the hybrid model has the best performance.

The optimization of software testing is one of the essential problems. In this paper, a stochastic Markov Decision Process (MDP) model of software testing is proposed, and the process of software testing is described as a reinforcement learning problem. A learning strategy based on the policy iteration of dynamic programming is presented to obtain the optimal testing profile. The case study indicates that, compared with random testing strategy, our learning strategy can significantly reduce the testing cost to detect and remove a certain number of software defects.