In this project we selected a heart stroke dataset and an email spam one for which we developed machine learning prediction systems by using a number of well-known data mining techiques like data preprocessing,data cleaning,classification using predictors like linear regression,KNN,Random Forests and advanced one's like neural networks.In the case of the heart stroke dataset , imbalance was detected in the class-goal of the prediction, namely the class with the name "stroke" where the number of patients that didn't have a stroke far outweighed those that did.For that reason we used an oversampling technique named SMOTE to balance the dataset for that class so we could conduct a trustworthy and valuable prediction of stroke possibility which otherwise would have been impossible.