JSE_V9_N4_RP6 T. Vara Prasad C. Silpa A. Srinivasulu Journal on Software Engineering 2230–7168 9 4 32 39 Software Defect Prediction, Software Metrics, Ensemble Learning Models The present generation software testing plays a major role in defect predication. Software defect data includes redundancy, correlation, feature irrelevance and missing value. It is hard to ensure that the software is defective or nondefective. Software applications on day-to-day businesses activities and software attribute prediction such as effort estimation, maintainability, and defect and quality classification are growing interest from both academic and industry communities. Software defect predication using several methods, in that random forest and gradient boosting are effective. Even though the defect datasets contain incomplete or irrelevant features. The proposed system Average probability ensemble technique used to overcome those problems and gives high classification result to compare another method. Because it has integrated with three algorithms to use classification performance and it gives more accurate result in publicly-available software datasets. April - June 2015 Copyright © 2015 i-manager publications. All rights reserved. i-manager Publications http://www.imanagerpublications.com/Article.aspx?ArticleId=3533