Pattern 2 Satisfiability Analysis via Hybrid Artificial Bee Colony Algorithm as a Learning Algorithm
Mohd. Asyraf Mansor, Mohd Shareduwan Mohd Kasihmuddin, Siti Zulaikha Mohd Jamaluddin, Saratha Sathasivam · 2020
A systematic and optimal learning algorithm is an essential domain in pattern verification and constraint satisfaction logic, such as in Pattern 2 Satisfiability (P2SAT). P2SAT is a class of Boolean based satisfiability problem that focuses on generating the optimum pattern with respect to the restricted Boolean logic. The task required a dynamic searching method in order to facilitate the verification process of generating the global PSAT pattern. The existing hybrid model of Artificial Bee Colony Algorithm (ABC) and discrete Hopfield Neural Network (DHNN) only focused in 2SAT logic programming with simulated data. In this research, we propose the optimal hybrid model by employing the hybrid ABC with DHNN in P2SAT. In order to verify the capability of our proposed model, we compare by employing the Genetic Algorithm (GA) as the learning method in doing PSAT as the benchmark. The experimental results manifested the performance of the proposed model in learning and retrieving the PSAT patterns. The analysis for the performance is based on root mean square error (RMSE), mean absolute percentage error (MAPE), the Global P2SAT and the CPU time evaluations obtained from the simulation. The simulations performed on different hybrid models reveal the power of hybrid ABC with DHNN in verifying and generating global P2SAT for the higher-order patterns. This study provides new insight and approach in P2SAT, especially in enhancing the learning algorithm to be more robust and systematic.