The prediction software defect components are an economically important activity and so has received a good deal of attention. However, making sense of the many, and sometimes seemingly inconsistent, a result is difficult. To improve the performance of software defect prediction, this research proposed the mixture of genetic algorithm and bagging technique. The thesis contains two phase. The first phase is feature selection; the features are selected using genetic algorithm, the bagging technique is used for class imbalance problem. The second phase is defect prediction; Software defects are predicted using K-Means and an Expectation Maximization (EM) algorithm. K-Means is a simple and popular approach that is widely used to cluster/classify data. EM algorithm is known to be an appropriate optimization for finding compact clusters. EM guarantees elegant convergence. EM algorithm assigns an object to a cluster according to a weight representing the probability of membership. The proposed method is evaluated using the data sets from NASA metric data repository. The proposed method is evaluated based on evaluation measurement such as accuracy and error rate. The experimental results demonstrate that our approach outperforms other competing approaches.