A software defect prediction model is used to help identifying the defects of software module. Software attributes play an important role to build the model. The performance and effectiveness of a software defection model depends on the characteristics of various software attributes which can be used to predict whether a software module is defected or non-defected. Based on software, one or more software metrics can become irrelevant as those do not contain significant information or contains redundant information. If appropriate attributes are not used for defect prediction model, the performance of the model may decrease. Therefore, in order to improve the effectiveness and performance of defect prediction model, it is important to select proper attributes which can be used to build a good predictor model. In this study, we propose an attribute selection process for software defect prediction model. Our experiment shows that our proposed approach provides comparatively better set of attributes which increases the performance of defect prediction model. We compared our approach with two existing approaches using eight NASA data set and our approach showed up to 54% improvement in software defect prediction.