Deep learning and fuzzy rule-based hybrid fusion model for data classification

Abstract

Data mining is the promising field that attracted the industries to manage huge volumes of data. The most effective and challenging techniques of data mining is data classification. The main intension of this research is to design and develop a data classification strategy based on hybrid fusion model using the deep learning approach, Adaptive Lion Fuzzy System (ALFS), and Robust Grey wolf based Sine Cosine Algorithm based Fuzzy System (RGSCA-FS). The hybrid model consists of three phases: In the first phase, the data is classified using ALFS and the rule base of the fuzzy system is updated by optimally generating the rules using adaptive lion optimization (ALA) from the training data. The second step is the fuzzification process, which converts the scalar values in the training data into fuzzy values with the help of membership function, which is based on Adaptive Genetic Fuzzy System (AGFS). Finally, the classified score of data instances is determined using defuzzification process, which converts the linguistic variable into fuzzy score. In the second phase, the data is classified using Robust Grey wolf based Sine Cosine Algorithm based Fuzzy System (RGSCA-FS), which is used for selecting the optimal fuzzy rules. In the third phase, the data is classified using deep learning networks. The outputs from three phases are fused together using the hybrid fusion model for which the weighed fusion is employed. The performance of the system is validated using three different datasets that are available in UCI machine learning repository. The proposed hybrid model outperforms the existing methods with sensitivity of 0.99, specificity of 0.9350, and accuracy of 0.9411, respectively.