Healthcare monitoring and Recommendation model: Hybrid Deep learning Enabled Disease Diagnosis Framework

Abstract

The continuous observation of patients’ health condition as well as monitoring varying crucial symptoms is termed as Healthcare monitoring, which also detect and trace any probable problems. This kind of observation is more crucial for detecting health conditions at early stage especially, recognizing cardiovascular disease in diabetes patient. However, diagnosing cardiovascular disease in diabetes patient is extremely difficult because of the small quantity of labeled data with the prevalence of outliers in the diabetes datasets. This paper proposes a novel healthcare monitoring and recommendation model with hybrid DL based disease diagnosis structure considering the heart disease. The improved z-score normalization is employed to preprocess the input data by scaling the numeric features within a general range. The BWSV-FE based feature extraction process is proposed to extract the entropy features; together with this feature, a raw feature and mutual information features are collected. Subsequently, the hybrid disease prediction model is proposed that includes improved Bi-GRU and LSTM models. Improved Bi-GRU is adopted for efficiently predicting the disease, in which the CCE is updated with the modified MSE loss function by inducing penalty function into this. Moreover, the penalty is computed by employing Gini Index function. Thereby, the hybrid disease prediction model efficiently predicts the diseases’ severity level that aids the doctor for guiding the management of the patients’ health.