In contemporary computer systems, accurately predicting the direction of program execution at decision points (conditional branches) is essential for enhancing processor performance. This paper introduces a novel approach leveraging machine learning tools, specifically Support Vector Machine (SVM) and Decision Tree algorithms, to improve branch prediction accuracy. Our approach consists of a systematic approach to training SVM and Decision Tree models using historical data with respect to the behavioral history of branches in past executions, which allows the models to learn the patterns contained in the relationship of program instructions and the respective decision outcomes. SVM addresses complex, non-linear relationships, whereas Decision Tree does relatively simpler ones, making our simulator suitable for different types of programs with different execution patterns. The simulator's effectiveness was established on a dataset with assessment metrics including prediction accuracy, error rates and instruction processing time. Our machine learning approach specifically added benefits in predicting complex and non-linear patterns of branch structures which has not been the case in traditional approaches. The results indicate that the performance of SVM and Decision Tree algorithms can be used in combination with branch prediction, where the timing constraints can be lessened. This machine learning based simulator sets new horizons in branch prediction bringing new fields into computer architecture design. This work helps to maintain the continuous flow of research in the direction of increasing the efficiency of processors by means of new approaches to branch prediction in which architecture de-coupling is performed.