Multi Disease Prediction Using Deep Learning Framework for Electric Health Record

Abstract

Information mining for medical services is an interdisciplinary field of study that started in data set measurements and
is helpful in looking at the adequacy of clinical treatments. For minor side effects, the trouble is to meet the specialists
whenever in the medical clinic. Hence, huge information gives fundamental information with respect to the disease
based on the patient's side effects. Thereby, prediction multiple disease with respect to effective energy technology as
deep learning. The stages of this paper's effective multi-disease prediction are as follows: a) Gathering data from
disease repositories, such as those for diabetes, heart disease, Alzheimer's disease, and pancreatic cancer;
b) Preprocessing using grey scaling, histogram equalization, and data normalization; c) Using a convolution
autoencoder to extract features; d) using particle swarm optimization (PSO) to select features; and e) using a recurrent
neural network (RNN) for prediction. Several state-of-the-art models are evaluated experimentally using a variety of
metrics, and the suggested model performs better (Diabetes; precision: 0.95; heart disease; precision: 0.93;
Alzheimer's disease; precision: 0.9; pancreatic cancer; precision: 0.93).