Evaluation and Improving Hypergraph Based Learning Algorithm over Data Integration Problem for Cancer Related Genes

Reliable predictive model build using semi supervised learning utilising classification algorithm has evolved rapidly in successful cancer treatment. In order to optimise the data integration problem, such as hypergraph based learning to integrate microarray gene expressions and protein interactions for predicting cancer outcome, novice optimization techniques are employed. The need of the hour is to have a good optimisation technique that would converge within acceptable amount of time in predicting promising result. So we propose an optimisation technique that is used in the first step of two step iterative method that alternatively optimises the labelling of samples for the hypergraph based learning. This learning method incorporates gene interactions as prior knowledge in protein interaction network. This optimisation technique can be imposed in various learning algorithm which utilises the principle iteration for optimisation. The proposed solution using Gauss-Seidel method converges faster and has better time complexity.