RBFN Networks-Based Models for Estimating Software Development Effort: A Cross-Validation Study

Software effort estimation is very crucial and there is always a need to improve its accuracy as much as possible. Several estimation techniques have been developed in this regard and it is difficult to determine which model gives more accurate estimation on which dataset. Among all proposed methods, the Radial Basis Function Neural (RBFN) networks models have presented promising results in software effort estimation. The main objective of this research is to evaluate the RBFN networks construction based on both hard and fuzzy C-means clustering algorithms using cross-validation approach. The objective of this replication study is to investigate if the RBFN-based models learned from the training data are able to estimate accurately the efforts of yet unseen data. This evaluation uses two historical datasets, namely COCOMO81 and ISBSG R8.