A modified reverse-based analysis logic mining model with Weighted Random 2 Satisfiability logic in Discrete Hopfield Neural Network and multi-objective training of Modified Niched Genetic Algorithm
Nur Ezlin Zamri,Mohd. Asyraf Mansor,Mohd Shareduwan Mohd Kasihmuddin,Siti Syatirah Sidik,Alyaa Alway,Nurul Atiqah Romli,Yueling Guo,Siti Zulaikha Mohd Jamaludin
DOI: https://doi.org/10.1016/j.eswa.2023.122307
IF: 8.5
2024-04-01
Expert Systems with Applications
Abstract:Over the years, the study on logic mining approach has increased exponentially. However, most logic mining models disregarded any efforts in expanding the search space which led to poor generalizability property of the retrieved induced logic. In light of this gap, this paper initiated the hybridization of logic mining approach with a multi-objective training algorithm namely Modified Niche Genetic Algorithm. The core impetus of this algorithm is to ensure optimal production of multiple superstrings via Wan Abdullah method resulting in multiple units associative memory feature of the Discrete Hopfield Neural Network. Therefore, the storage capacity of DHNN increases which directed towards larger search space of locating optimal induced logic. Additionally, several modifications were imposed to counter other issues such as, rigid logical rule, outdated quality of best logic, and high dependency on the supervised attributes selection method. Experimentation was done on 20 repository datasets from reputable machine learning repositories. Results showed that the proposed model outperformed all baseline methods in terms of accuracy = 0.8727, precision = 0.9845, specificity = 0.9988, and Matthew’s correlation coefficient = 0.5815.
computer science, artificial intelligence,engineering, electrical & electronic,operations research & management science