Applying Fuzzy Clustering To Predict Software Usability

Estimation of different software parameters based on measured attributes from previous similar products is an active field of research for developers as well as managers. Such estimation models must inevitably handle imprecision and uncertainty and hence soft computing techniques are gaining popularity. Fuzzy logic integrated with data mining techniques becomes one of the key constituents of soft computing in handling the challenges posed by massive collections of natural data. In this paper Fuzzy C-Means algorithm is applied to a set of Software Usability data and clusters are generated. These clusters centroids are used to generate the initial set of fuzzy rules which can be refined with the help of train data. These fuzzy rules can be used to label the clusters with the software usability values. With these labeled clusters, new unlabeled data can be allocated to clusters with fuzzy membership values. The advantage of fuzzy clustering is that even though we can allocate an absolute cluster membership to a data point, at a finer granularity level we can provide the percentage usability.