Risk is a destructive and serious problem that may happen in the future of the software project. The definition of risk is an indicated negative thing. However, in software project development risk has two meanings. It is the opportunity for the developer and customer if it gives motivational consequence on the project objectives and the other one is a negative risk that affects the project objectives and may lead to the failure of the project. Improper assessment of negative risk is the main reason for the failure of SP. In this thesis, we proposed enhanced software risk assessment and prediction using machine learning to overcome the problem. Proper risk assessment is critical for increasing organizational income, process improvement, programmatic optimization, and the Project manager can easily make decisions. In this study firstly comparison of classification models with ROC curve Area was analyzed and secondly comparison of models with evaluation metrics. From the models (Support vector machine (SVM), Random Forest (RF), and Logistic Regression (LR)) the classification performance of LR was shown superior on SVM and RF. The ROC curve area of LR was 1.00 for class 0 and the Macro average ROC curve area of LR is 0.99. Next to LR, the ROC curve area of SVM was found to be 1.00 for class 0 and the Macro average ROC curve area of SVM was 0.97. From models (SVM, LR, AND RF) the performance of RF is shown to be accurate than SVM and LR. The MAE and R-square of RF were 0.21 and 0.86 respectively. SVM follow with the MAE and R-square of 0.25 and 0.82 respectively.LR was the least performer with the MAE and R-square of 0.32 and 0.78 values.