Improved Harmony Search with Optimal Deep Learning Enabled Classification Model

Due to drastic increase in the generation of data, it is tedious to examine and derive high level knowledge from the data. The rising trends of high dimension data gathering and problem representation necessitates feature selection process in several machine learning processes. The feature selection procedure establishes a generally encountered issue of global combinatorial optimization. The FS process can lessen the number of features by the removal of unwanted and repetitive data. In this aspect, this article introduces an improved harmony search based global optimization for feature selection with optimal deep learning (IHSFS-ODL) enabled classification model. The proposed IHSFS-ODL technique intends to reduce the curse of dimensionality and enhance classification outcomes. In addition, the IHSFS-ODL technique derives an IHSFS technique by the use of local search method with traditional harmony search algorithm (HSA) for global optimization. Besides, ODL based classifier including quantum behaved particle swarm optimization (QPSO) with gated recurrent unit (GRU) is applied for data classification process. The utilization of HSA for the choice of features and QPSO algorithm for hyper parameter tuning processes helps to accomplish maximum classification performance. In order to demonstrate the enhanced outcomes of the IHSFS-ODL technique, a series of simulations were carried out and the results reported the betterment over its recent state of art approaches.