LSTM-Dropout Model in Software Reliability Prediction

Document Type : Original Article


1 Dept. of Algorithms and Computation, School of Eng. Sci., Col ledge of Eng. University of Tehran

2 Department of Algorithms and Computation, School of Engineering Sciences, College of Engineering, University of Tehran, Tehran, IRAN


There exist numerous methods which have been introduced to predict the reliability of a software. These methods may be, in general, partitioned into two main categories, namely parametric (e.g. software reliability growth models) and non-parametric (e.g. neural networks). Both approaches have been successfully implemented in software testing applications over the past four decades. Most of the time, the data in reliability prediction applications are provided in the form of time series. Due to this reason, deep recurrent networks models (e.g. LSTM network) are among the most powerful methods in reliability related problems. However, a major concern when using deep neural networks for software reliability applications, is the problem of overfitting. To address this issue, we make use of dropout and LSTM Encoder-Decoder methods which yield promising results.


Main Subjects

Articles in Press, Accepted Manuscript
Available Online from 07 March 2022
  • Receive Date: 21 September 2021
  • Revise Date: 30 October 2021
  • Accept Date: 07 March 2022
  • First Publish Date: 07 March 2022