Software Defect Prediction (SDP) is a method of classification that check software is defective or not. Software defect is some kind of flaw, error or some kind of mistake from the development team which prevent the software from the smooth working. SDP at the beginning stage of Software Development Life Cycle (SDLC) will generate some results which will help us to solve the problem and help us to reduce development cost. In this paper, we are comparing nine open source software which are written in java language from PROMISE repository, and for it we are using main four feature extraction techniques like Auto-encoders, Linear Discriminant Analysis (LDA), PrincipalComponent Analysis (PCA), Kernel-based Principal Component Analysis (K-PCA) with base classifiers of machine learning, Support Vector Machine (SVM). Ten-fold cross-validation method is used to perform model validation and model efficiency is calculated with the help of accuracy and ROC-AUC. The outcome of this study defines that Auto-encoders is a very effective technique to reduce the measures of defect dataset of software successfully.