An active study field in software engineering is software defect prediction. Before the testing phase even begins, the defect-prone modules are identified using the defect prediction approach. An effective defect prediction model utilizes a few software metrics in order to improve defect prediction. Metrics-based modules enhance software quality, cut costs, and enable efficient resource allocation. Employing classification of defect data with a classifier will increase the effectiveness of defect prediction. Software metrics like Halstead metrics, McCabe's metrics, and LOC based metrics of each module is measured and recorded as a dataset. In this study a real time software project dataset KC1 is taken from NASA Metrics Data Program. Naive Bayes algorithm, Support Vector Machine algorithm, K Nearest Neighbour algorithm and NB Simple algorithm are used as classifiers. The classifications performance is measured using Exactness, Accuracy, Recollection and F Measure. This paper concludes for the used defect dataset an accuracy of 98.89 is obtained with Support Vector Machine algorithm as the best classifier.