Multi Variant Deep Transfer Learning Analytics Model for Improved Health Care Solution Using Big Data

The recent surge in healthcare big data has presented
new possibilities of intelligent disease prediction, but currently
used analytics models fail to handle heterogeneous data, feature
imbalance, and are not generalized across clinical domains. This
paper tries to resolve these issues with a new Multi Variant Deep
Transfer Analytics Model (DVDTAM) of better healthcare
decision support. The aim is to improve the accuracy of disease
prediction and reduce the rates of false classification. DVDTAM
has a Deep Variant Feature Normalizer that uses the Feature
Covariance Value to stabilize heterogeneous distributions of
features after which a Multi Variant Deep Transformer with
transfer learning is used to model complex clinical dependencies.
Final disease classification is done using a Multi Factor Disease
Absorption Weight mechanism. As it is experimentally assessed
on the MIMIC-III dataset, DVDTAM reaches a high accuracy of
96.8% and is better than state-of-the-art models by up to 2.5
percent, and the false rate is decreased to 0.05. These findings
support the usefulness of the suggested structure of scalable and
dependable healthcare analytics.