OFES: Optimal feature evaluation and selection for multi-class classification

The complexity and accuracy of classification algorithms largely depend on the size quality feature set used to build classifiers. Feature evaluation selection are critical steps decide a small high-quality features accurate efficient classifiers since low-quality not only have negative impacts results but also increase algorithms. Current popular sufficient in selecting discarding features, especially for streaming data. This paper proposes novel approach, optimal (OFES), evaluate select multi-class classification. OFES first measures difference between any two classes based that is be evaluated. Then, it defines quantitative identify features. Applying application identifies users their arm movement patterns, we find when compared with other approaches, such as Information Gain Ranking Random Projections Matlab ranking, improves regardless different It demonstrates great scalability number yields higher 95%.