AutoFusion of Feature Pruning via Disciplined Convex Concave Programming

Fusion of algorithms has recently taken attention in machine learning studies because of its power coming from community decision instead of single decision maker. One of the crucial questions in aggregation of algorithms is that which and how many models should be combined to achieve both best accuracy and low complexity. It is already knownn in machine learning that as the complexity of the model increases too much, prediction accuracy decreases. There is a trade of between these two features. In order to answer such questions, diversity notion gets involved to overall consensus models. It is also shown that diversity alone does not determine the best ensemble (fusion), so accuracy and diversity together has been taken into account recently in such problems. In this paper, we took in account of those two notions simultaneously so that number of algorithms and which algorithms should be in the ensemble are answered while solving feature selection problem. We have validated our algorithm on different domains of data sets which shows better prediction accuracy values than existing ensemble based feature selection methods.