Software Reliability Predictions Research Articles

Failure prediction by regularized fuzzy learning with intelligent parameters selection

Software failure time prediction, which is critical in software dependable evaluation, is among one of the most critical issues in software quality assurance research. Conventional software reliability predictions often use single models to illustrate the overall characteristic of the process of software failure. However, those models often fail to produce good prediction performance owing to the constant changes in software failure patterns during different time periods. Fuzzy systems like Takagi–Sugeno–Kang are multimodel approaches that combine simple submodels to represent the overall characteristic of dynamic nonlinear complex systems. This paper presents a novel Fuzzy system based software reliability prediction model, called regularized extreme learning adaptive neuro-fuzzy inference system, and employs the quantum-inspired binary gravitational search algorithm to determine its suitable regularization parameters. The new model can effectively reduce randomness, computational complexity as well as avoid local optimization when searching for optimal parameters. An experimentation using four real-world software failure data sets was carried out to compare the performance of the proposed model with that of certain representative software reliability prediction models. The results showed that the proposed model with the optimized parameters exhibits superior performance.

Reliability assessment of service-based software under operational profile uncertainty

We address the problem of operational reliability assessment through testing of software services delivered on-demand such as Web Services. Software reliability assessment is typically done for a specific operational profile: the profile is needed in testing to select or generate test cases (operational testing) in a way statistically similar to the anticipated use of software in operation; the observations of success/failure of test executions are used to predict software reliability in actual operation. It is well known that unless the profile is accurate, software reliability predictions obtained via operational testing cannot be trusted.We present a new way of dealing with the uncertainty in the operational profile adopting a two-stage Bayesian inference for reliability assessment. The technique relies on the availability of information about partitions of the input space. The approach is demonstrated on contrived examples and on a case study of real Web Services. We discuss the usefulness of the approach in dealing with two important practical problems: i) the true profile in operation differs from the one used in testing, ii) the profile in operation is changing continuously.

Novel Algorithms for Web Software Fault Prediction

Reliability is gaining importance with time for Web system because of the popularity of different Web‐based applications, namely, different Web sites and social community networks. In order to aggrandize the reliability of a Web system, some methods are required to measure its current reliability. In recent years, some research works have been carried out on Web software error analysis and reliability predictions. In all these research works, the Web environment has been considered as a crisp one. This is not in reality. Hence, in this paper, a novel clustering algorithm and a multivariate fuzzy logic and fuzzy time series based prediction algorithm for Web software fault prediction have been developed. The proposed prediction algorithm can predict the occurrences of more than one Web error, in a single day, on a single run. Proposed methods have been validated using some real Web failure data extracted from the HTTP logs (access and error logs) of www.ismdhanbad.ac.in (the official Web site of Indian School of Mines Dhanbad, India), which were collected from the Indian School of Mines Web server. Copyright © 2014 John Wiley & Sons, Ltd.

Web software fault prediction under fuzzy environment using MODULO-M multivariate overlapping fuzzy clustering algorithm and newly proposed revised prediction algorithm

In recent years some research works have been carried out on web software error analysis and reliability predictions. In all these works the web environment has been considered as crisp one, which is not a very realistic assumption. Moreover, web error forecasting remains unworthy for the researchers for quite a long time. Furthermore, among various well known forecasting techniques, fuzzy time series based methods are extensively used, though they are suffering from some serious drawbacks, viz., fixed sized intervals, using some fixed membership values (0, 0.5, and 1) and moreover, the defuzzification process only deals with the factor that is to be predicted. Prompted by these facts, the present authors have proposed a novel multivariate fuzzy forecasting algorithm that is able to remove all the aforementioned drawbacks as also can predict the future occurrences of different web failures (considering the web environment as fuzzy) with better predictive accuracy. Also, the complexity analysis of the proposed algorithm is done to unveil its better run time complexity. Moreover, the comparisons with the other existing frequently used forecasting algorithms were performed to demonstrate its better efficiency and predictive accuracy. Additionally, at the very end, the developed algorithm was applied on the real web failure data of http://www.ismdhanbad.ac.in/, the official website of ISM Dhanbad, India, collected from the corresponding HTTP log files.

Comparative analysis of software reliability predictions using statistical and machine learning methods

This paper examines the performance of statistical linear regression and machine learning methods like Radial Basis Function Network RBFN, Generalised Regression Neural Network GRNN, Support Vector Machine SVM, Fuzzy Inference System FIS, Adaptive Neuro Fuzzy Inference System ANFIS, Gene Expression Programming GEP, Group Method of Data Handling GMDH and Multivariate Adaptive Regression Splines MARS for predicting software reliability. The effectiveness of LR and machine learning methods are illustrated with the help of 16 failure datasets of real-life projects taken from Data and Analysis Centre for Software DACS. Two performance measures, Root Mean Squared Error RMSE and Mean Absolute Percentage Error MAPE, are compared quantitatively obtained from rigours experiments. We empirically demonstrate that performance of the SVM model is better than LR and other machine learning techniques in all datasets. Finally, we conclude that such methods can help in reliability prediction using real-life failure datasets.

Combining Architecture-based Software Reliability Predictions with Financial Impact Calculations

Software failures can lead to substantial costs for the user. Existing models for software reliability prediction do not provide much insight into this financial impact. Our approach presents a first step towards the integration of reliability prediction from the IT perspective and the business perspective. We show that failure impact should be taken into account not only at their date of occurrence but already in the design stage of the development. First we model cost relevant business processes as well as the associated IT layer and then connect them to failure probabilities. Based on this we conduct a reliability and cost estimation. The method is illustrated by a case study.

Finding the optimal parameters for a software reliability model

Failure rate parameters exert a great deal of influence on the accuracy of software reliability predictions. Thus, it behooves us to examine the methods that are used to estimate the parameters. The value of parameters is determined by the pattern of historical failure data and the amount of data used in the estimates. Conventional wisdom suggests that all the data be used. However, this could be counterproductive if all the data are not representative of the current and future failure process. Therefore, we can analyze these factors to obtain the optimal parameter estimates, where “optimal” means parameter values that will result in minimum prediction error. We examine one software reliability model that does not use all the failure data if a priori analysis indicates that optimal parameter values could not be obtained. Historically, this model has used a mean square error criterion (MSE) of the difference between actual and predicted reliability in the parameter estimation range. However, it has been observed that this policy does not always result in minimum error predictions in the prediction range. Therefore, we investigated the linear interpolation estimation method as an alternative to MSE for three NASA shuttle flight software releases comprising forty-two reliability predictions. Interestingly, linear interpolation provided better prediction accuracy, overall. We believe our analysis can be applied to other software and failure data. In order to systematize our thinking on developing new models using linear interpolation, we employed object-oriented design concepts as guides to implementing C++ programs for evaluating linear interpolation with respect to MSE.

Why software reliability predictions fail

Software reliability reflects a customer's view of the products we build and test, as it is usually measured in terms of failures experienced during regular system use. But our testing strategy is often based on early product measures, since we cannot measure failures until the software is placed in the field. The author shows us that such measurement is not effective at predicting the likely reliability of the delivered software.

Reliability prediction and estimation of PROLOG programs

This paper presents an approach to reliability prediction and estimation of PROLOG programs, and introduces 2 complexity measures for PROLOG programs. The structural complexity measure refers to the program's static characteristics: size and number of clauses; clause arguments (number and types); and clause types (facts or rules). The operational complexity measure refers to the program's dynamic characteristics: execution frequency of program components; user behavior; and backtracking and recursion. Values of the two measures are used to: (1) predict PROLOG program reliability before testing and in the early testing stages; and (2) estimate the reliability as a function of time, in order to determine whether the reliability objective is achieved. The feature-oriented reliability determination approach leads to improvements in the accuracy of software reliability predictions and estimations. >

136 Integration of software reliability predictions to achieve modelling fault tolerance

136 Integration of software reliability predictions to achieve modelling fault tolerance

Integration of Software Reliability Predictions to Achieve Modelling Fault Tolerance

A considerable number of reliability growth models has been proposed to evaluate software dependability. Each of them provides an inference rule from past failure data on the present and future behaviour; they are based on different assumptions, some of which are only realistic for particular application classes and unprovable at present. Also for this reason none of the models has ever shown to be the absolutely best: its accuracy varies from an application to another and cannot be predicted a priori. Although it may be easy to identify the worst models in a given case by means of absolute quality measures quantifying the departure of prediction from reality, the identification of the best one is particularly difficult, as the goodness of fit values may be time-variant. Thus even an experienced model user may encounter considerable difficulties in making up his mind when selecting the most promising modelization for the specific situation considered. In order to simplify the decision phase, allowing the use of evaluating tools also to appliers non-familiar with the underlying modelling theories, we suggest to automatize the model choice, by selecting deterministically the estimation value considered to be most reliable. The intention of this contribution is to study two particular strategies integrating different predictions, which have been suggested in literature, in particular with respect to their sensitivity to model-behavioural changes, identifying the most critical and most suitable situations to be treated by each of them.

Evaluation of competing software reliability predictions

Different software reliability models can produce very different answers when called on to predict future reliability in a reliability growth context. Users need to know which, if any, of the competing predictions are trustworthy. Some techniques are presented which form the basis of a partial solution to this problem. Rather than attempting to decide which model is generally best, the approach adopted allows a user to decide on the most appropriate model for each application.

Software Reliability Predictions

The authors have developed three practical models for prediction of software reliability based on data consisting of the times between error occurrences. In the current application, the models employ data consisting of a record of the number of errors versus CPU time used in testing of a program. The models and their applications are discussed in Part I.