Disease prediction by using deep learning based on patient treatment history

Abstract

Now a day’s medical and healthcare industries has big business. Healthcare industry produces large amount of data in daily rou-tines. That big amount of data is used for the future disease prediction. Prediction is does on the patient previous history and health related information. In every hospital or clinics there are large amount of patient history or patient related all information is available. But there is the major challenge is that how to extract that information from that large data records. For doing the prediction of diseases using that patient treatment history by applying the machine learning and the data mining techniques is the continuous struggle for the past few years in medical or healthcare industry. In previous years many papers are published by using the data mining techniques and machine learning technique for the disease prediction, also for the progression and reoccurrence of those diseases. In this paper we build the new model for the diseases prediction. In that we use the deep learning concept artificial neural network (ANN) for predicting the diseases. In this paper we use the probabilistic modelling and deep learning concept for prediction. For that we collect the three diseases heart, kidney, and diabetic’s dataset. For those diseases we build the one proper dataset. That dataset are split into the training and testing dataset. For training the dataset we use the scholastic gradient decent algorithm. For this project we collect patient related information. We collect the datasets from the UCI Repository, Pima datasets, and Kaggle datasets. For that collected dataset we apply the pre-processing and remove the unnecessary data and extract the important features from that data. On the new generated data we apply the probabilistic model and deep learning technique for doing the diseases prediction. Then by using ANN method we do the prediction and produce the confusion matrix. Then trained and tested model will be deployed in a real-life scenario for diseases prediction. We got the prediction accuracy 95%, 98%, 72% for heart, kidney, and diabetes disease respectively which is far higher when compared to the existing methods.