Margin based feature selection: An algorithmic approach for a set of attributes extrication

Abstract

In information processing industry, mining techniques plays a major role to analyze the huge dataset. Information/Data-Mining is a extraction-progression meaningful information from the dataset. The major steps in data mining are pre-processing, mining and result validation. Classification is one of an important methodology to detect disease for the humans. There are lots of algorithms for classification. Feature selection is a useful method for classification and clustering. This work aims to classify disease based on health information and to take treatment on the early stage with novel feature selection algorithm and ensemble learning based classification. This will improve the feature selection accuracy. Breast cancer detection application is an example of this type of classification. Classification and prediction problems have a vital role in medical decision making. Disease diagnosis is a multiclass classification problem. The classification in disease diagnosis is t o assign a disease label to a particular instance. High dimensional datasets have the problem of presence of unrelated otherwise superfluous features, which often lowers the performance of machine learning algorithm. A suitable feature selection method is required for high dimensional data set classification. I n t his paper, feature selection algorithm named as Margin based Characteristic Analysis procedures and data combining methodologies are going to use.