Faults Prediction for Component Based Software using Interrelated Feed Forward Neural Networks

The software reliability generally depends upon the rate of failures, or the number of faults occurred in mean execution time and the numbers of faults are expected to occur during the course of execution of software. There are various models have been proposed to predict the expected number of faults for various types of software structure. In this paper we are predicting the number of expected faults occurred in each component of software as well as for the whole software in the expected execution time interval using component based neural network architecture. The simulation design and implementation results are suggesting there may be more number of predicted faults in components than the number of predicted faults in the complete software.