A Meta-Heuristic Approach for Optimizing Neural Network Model for Heart Disease Prediction

In heart-disease detection, we have several systems utilizing Machine Learning models to assess the likelihood of a person experiencing heart disease based on input data gathered from health checkups. To deploy these in real-time, we need to constantly update the model with validation data generated...

Full description

Saved in:
Bibliographic Details
Published in:Proceedings of the XXth Conference of Open Innovations Association FRUCT 2024-04, Vol.35 (2), p.830-https://youtu.be/tty6CT_RgAA
Main Authors: Gummuluri Venkata Ravi Ram, Jayanth Prathipati, Hemanth Kumar Mogilipalem, Sona Mundody, Ram Mohan Reddy Guddeti
Format: Article
Language:English
Subjects:
ant colony
artificial neural networks
genetic
heart-disease
hyperparameter
meta-heuristic algorithms
swarm-bee
tuning
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In heart-disease detection, we have several systems utilizing Machine Learning models to assess the likelihood of a person experiencing heart disease based on input data gathered from health checkups. To deploy these in real-time, we need to constantly update the model with validation data generated from the feedback of doctors and users. One crucial component is hyperparameter tuning, as hyperparameters define the working conditions of a model. Current hyperparameter tuning methods have more time-complexity, which might cause latency. Meta-heuristic algorithms are known to achieve optimal solutions in less time, leveraging accuracy. In this paper, we performed time-complexity analysis for three bio-inspired meta-heuristic algorithms viz. Genetic Algorithm, Ant-Colony Optimization and Swarm-bee optimization for hyperparameter tuning in the context of Artificial Neural Networks in heart disease prediction. We achieved accuracy 93.17 on Cleveland dataset and 95.12 on Cleveland, Hungary, Switzerland, and Long Beach V combined dataset, which outperformed conventional algorithms in less time complexity. We built a web-app and mobile-app based on best models.
ISSN:2305-7254
2343-0737
DOI:10.5281/zenodo.11096943