Recently there are a lot of advancements in healthcare technology. Amongst, Electronic Health Record (EHR) is an upcoming trend which stores patients’ demographics, lab tests & results, medical history, habits etc. collaborated in electronic form. EHR is huge data, which is difficult to maintain and retrieve. So the idea of health risk prediction is formulated in this work. To get the relevant data from EHR, feature selection technique is used. Feature selection is responsible to collect only important and needed data from the dataset. For feature selection regression method is used in which loss function is proposed due to which accuracy and performance of the model are increased. Further risk prediction is done using neural network model. Deep Neural Network (DNN) is best suited for pattern learning and prediction purpose. It consists of various layers which have their specific function. DNN uses transfer learning to avoid repeated training for the whole system. Dataset considered here is of hypertension. EHR data is also synthetically created for analysis.