An Early Software Reliability Prediction Method using Backpropagation Algorithm

The objective of this paper is to predict software reliability based on software complexity metrics before the software testing phase. After finding the attributes which influence on the software reliability about ATM switching system, we define the software complexity of CHILL language based on them. We use the backpropagation algorithm of artificial neural networks in order to predict the fault proness of a function block. By predicting the software reliability at the implementation phase, we can recognize the unreliable function blocks in the software parts efficiently. Also, this prediction result can be used as a qualitative basis to decide whether a redesign is necessary or not.