| dc.description.abstract | Artificial Intelligence and Machine Learning has become the biggest phenomena in the last decade in the world, especially after the emergence of the Deep Learning paradigm then followed by the development of Reinforcement Learning. However, machine learning still has limitations both in terms of accuracy for minimal data and effectiveness. Therefore, we propose two new methods, namely NeuroGenGPR and NeuroGenGPR-RL. Where the NeuroGenGPR method comes from the Neuro-Genetic Approach combined with Gaussian Process Regression with Multi-Paired Whale Optimization Algorithm optimization, while the new NeuroGenGPR-RL method is an elaboration of the Neuro-Genetic Approach with Gaussian Process Regression and Multi-Paired Whale Optimization Algorithm optimization. in a Reinforcement Learning environment. These two new models are expected to be able to solve problems in prediction and classification problems with minimal data. Based on research conducted, NeuroGenGPR produces better prediction results compared to LSTM, GRU and GPR-WOA. NeuroGenGPR on Batubara Regency data produces RMSE of 3.6113 and 2.7440 for data ratios of 8:2 and 9:1, while LSTM produces 4.0793 and 3.5809, GRU produces 4.0646 and 3.5328, and GPR-WOA produces 5.5834 and 4.1679. Meanwhile, for Asahan Regency, NeuroGenGPR produces RMSE of 4.8772 and 4.7439 for data ratios of 8:2 and 9:1, while LSTM produces RMSE of 5.0207 and 5.0231, GRU produces 5.6836 and 4.9843, and GPR-WOA produces 8.2568 and 7.4689. For Labura Regency, NeuroGenGPR produces RMSE of 2.9077 and 2.3844 for data ratios of 8:2 and 9:1. Meanwhile, LSTM produces RMSE of 3.5396 and 3.7975, GRU of 3.9345 and 3.7550, while GPR-WOA produces RMSE of 4.0764 and 4.0473. Likewise, NeuroGenGPR-RL produces better classification results in terms of accuracy compared to the LSTM and GPRC models. NeuroGen GPR produced an accuracy of 99.28% compared to LSTM with 99.00% and GRPC with 98.29%. | en_US |
