As software systems become larger and more complicated, the task of detecting and fixing defects to improve the software performance is getting duller and inefficient. Developers employ issue tracking systems to collect defects for software improvement. Users submit defects through such issue tracking systems and decide the severity and priority of reported defects. It helps developers to solve important defects on time. However, manual severity and priority assessment is a tedious job and could be incorrect. Various researches have been done to automate detection of defect severity and priority in separate manner. Almost all studies were concentrated on detection of defcts only. Automatic detection and predicting of severity and priority was neglected in previous studies. Thus, in this work, we present a system for automatic detection and predicting of software defects severity and priority by using previous bugzilla defect reports.The proposed model has three components: preprocessing,feature extraction and classification. A 5-way Softmax is used for predicting into a specific class of defect severity and priority (minor,normal,major,critical,blocker and p1,p2,p3,p4,p5) respectively.The proposed model is implemented using matlab2019b (using recent version of matlab) and tested using sample defect report texts dataset collected from different bug tracking systems (bugzilla of eclipse,genome,firefox). The model achieved a prediction accuracy of 91.03% for training and 90.05% for testing to detect and predict software defect severity and priority. Our model was faster to train and had smaller model size as compared to the machine learning. In addition, the application of LSTM and word2vector are also used to improve the performance of state-of-the-art models as well by 4% (datamining and machine learning ) and 3% (ANN) respectively.