Artificial Neural Network based Model for Reliability Assessment of Component based Software

Software quality assessment during early phases of software development process is an extremely important concern of the researchers today because it identifies various aspects of quality degradation much prior to doing the actual damage to the final product quality. This also serves as the basis for improvement of the process for developing software. Software reliability is one of prime factor that affects software quality. In this paper, a model based on Artificial Neural Network (ANN) for the assessment of reliability of Component Based Software Systems (CBSS) has been proposed. First, a mathematical model based on formulation of software reliability in terms of the reliability factors using Analytical Hierarchy Process (AHP) is developed. Further, this model is refined by using ANN to calculate appropriate weight values reflecting true influence of reliability factors on software reliability. Model is validated by assessing the quality of 100 component based software using proposed models and an existing model. Results show a good correlation between the mathematical model and the ANN model. The proposed AHP model and ANN model achieve higher average reliability value of 0.5109 and 0.5088 respectively in comparison to the average reliability value of existing software reliability model (0.2261). Precise assessment of software reliability through the proposed models during early stages of software development helps developers to improve quality of software.