Development of Augmented Deep Learning Approach for Predicting and Diagnosing Generic Health Issues

Abstract

In the current context of epidemics, people are increasingly afraid to seek hospitalization. Moreover, dealing with all patients on the physician side is a hugely challenging subject. Hence our research work trying to develop an automated naval framework, which automatically detect abnormal patient from the set of all the patient, and allow those specific patients only to the medical expertise knowledge. Artificial Intelligence based platform is used to accelerate the progress for diagnosing and providing treatment for various diseases. Various deep learning methodologies have been glorified to reach its goal using integrated bio informatics approaches. The data collection is improved with the help of big data analytics which supports cluster computing and automated data processing. So, it becomes easier not only for collecting such kind of data but also for producing the comprehensive healthcare report by converting them into a critical insight. Then Word2vec with Recurrent conditional random filed is utilised in order to extract information. This hybrid data driven findings is interpreted to recognize patterns and analyses the structure to the finest level. So, it helps the doctor to diagnose specific diseases faster and more accurate. From the extracted information, the disease can be easily recognized and diagnosed with the help of combination of NER (Named Entity Recognition) and Bi-directional LSTM (Long short-term memory) algorithms. These algorithms not only identify the keywords but also the context of entities which combines direct matching and stemmed matching. So, it can able to process the health care information based on the lifestyle and environment. Finally, we compare our proposed augmented technique against various state-of-art model, as a result our augmented technique produces better performance against state-of-art models in terms of Accuracy, Recall, Precision and AUC.