Identifying fault-prone function blocks using the neural networks - an empirical study

An empirical study was carried out at ETRI to investigate the relationship between several metrics and the number of change requests associated with function blocks. For this, we propose the software metrics of CHILL language, which are used to develop the telecommunication software. We present the identification method of fault-prone function blocks of telecommunication software based on neural-networks. We focus on investigate the relationship between the proposed metrics and function block's faults which are found during development phase. Using the neural network model, we classify function blocks as either fault-prone or not fault prone based on the proposed metrics. Therefore, for newly added function blocks, we can predict whether the function block is fault-prone or not fault prone.