Cardiovascular diseases (CVDs) are a top cause of death around the world, leading to nearly 17.9 million deaths each year. Standard diagnostic instruments cannot identify risk factors for heart disease in their early development stages; therefore, nowadays medical solutions have failed to resolve this issue. Research looks at a wide range of studies that create heart disease models using traditional artificial intelligence methods and newer learning techniques, as well as combinations of both. Machine learning technology generates valuable mathematical systems that combine heart condition detection with the assessment of risk elements at their initial stages. This research uses Multilayer Perceptron (MLP) and k-Nearest Neighbors (kNN) methods. Feature selection methods optimize the choice of model variables, leading to the development of simpler, more readable models. The assessment evaluated both Cleveland and Framingham heart disease datasets through the accuracy and efficiency of heart disease models. Experimental results demonstrated that MLP projection outperformed all the tested techniques. MLP gives 89.97% accuracy, which is the highest among SVM and KNN methods. Eventually, it is stated that MLP is the best approach for predicting heart diseases by considering the lifestyle parameters.