The deterministic and probabilistic models are commonly used in software reliability engineering practice, and most of the decision-making are based on the results obtained from the analysis of selected model. If the selected model does not fit the collected software testing data relatively well, we would expect a low perdition ability of this model and the decision-making would be far from what is consider to be optimal decision. The two parameter Weibull distribution is proposed in this thesis as a probabilistic software reliability model. The shape and scale parameters of the Weibull model are estimated by using some estimation methods (MLE, LS, TTT). The validation and selection of this model are evaluated by using the empirical Q on test ratio (EQTR). A comparison between our suggested model (Weibull model) and a classical software reliability model (Schick-Wolverton) are implemented. According to the above comparison, our model is better than the classical model and thus can be used as a predictable software reliability model.