Incremental model of relevance in features with healthcare data analytics

With the rapid assembly of big data health frameworks for early lung cancer detection, this research provides solution for early and optimum diagnosis of lung cancer with high dimensional dataset by using, the extraction of features carried out with map-reduce framework followed by feature selection using correlation based model and relevance model. The model, in addition develops a Symmetric uncertainty and the Minimum Redundancy Maximum Relevancy (MRMR) model in combination with the weight computation model and incremental selection model. We advocate the use of the Map Reduce framework for processing large datasets that is to be used for classification. The results show the importance of our model with respect to the accuracy of the proposed model which has been increased and became stable at 91%, 92.04%, and 91.6% respectively for all high dimensional unstructured dataset Sputum images, Lung cancer, and thoracic datasets respectively. The error rate decreased to 8% for the lung cancer dataset, 9.5% for the sputum dataset, and 6.2% for the thoracic surgery dataset. Redundancy has been reduced as well as relevancy among features has been increased, 0.852 for sputum, 0.87 for lung cancer, and 0.90 for thoracic surgery datasets. Since it was able to eliminate redundancy and relevancy among several features, we also observed a reduced error rate in most of the datasets.