Skip to content
Conference Open access

Adaptive Feature Dependent Deep Transfer Learning Model for Improved Disease Prediction Using Big Data

Sep 2026 · 2026 International Conference on Secure IoT and Cybersecurity (ICOSICS) · pp. 1957-1962 · 0 citations · 17 references

Abstract

The use of big data and deep learning methods in healthcare analytics is becoming more important in terms of effective disease prediction in an early stage. Nevertheless, the current methods are prone to low performance because of incomplete, heterogeneous, and inconsistent clinical data that are gathered through various sources. The proposed research will solve these problems by proposing an Adaptive Feature Dependent Deep Transfer Learning Model (AFDDTM) to achieve more accurate disease prediction in a healthcare big data setting. The suggested framework adds an Adaptive Feature Mean Transformer (AFMT) to improve the completeness of data by calculating Feature Value Mean (FVM) and Feature Categorical Value (FCV), which makes features represented homogeneously. The use of dependency-aware deep transfer learning to capture latent feature correlations and in the disease-specific stratification and influence-based decision modeling to enhance class discrimination is also used. In the experimental analysis performed on a benchmark healthcare dataset it can be seen that AFDDTM has a prediction accuracy of 93.87 whereas other state of the art algorithms have a prediction accuracy of 96, this is because the false prediction rate at AFDDTM is a lot less at 6.13. The findings substantiate the performance of adaptive feature harmonization and influence-based learning in managing heterogeneous big data of healthcare. The suggested model is a valid and scalable clinical decision support system on the field.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.