Stroke is a major life threatening disease to cause of death and it has a serious long term disability. The time taken to recover from stroke disease depends on patient’s severity. Number of work has been carried out for predicting various diseases by comparing the performance of predictive data mining. In the proposed work MLFFNN with back propagation algorithm is used. The number of hidden neurons is optimized by genetic algorithm. This work demonstrates about ANN based prediction of stroke disease by improving the accuracy with higher consistent rat using optimized hidden neurons. In this algorithm determines the attributes involving more towards the prediction of stroke disease and predicts whether the patient is suffering from stroke disease. The data is collected from 300 patients. Among that 180 patients having disease .In the proposed work 196 data are used for training and 104 data is used to test the performance of the system.