Abstract

This paper presents the design and implementation of an efficient multiparametric recommender engine for low-power applications with incremental post-processor optimizations. The need for this work arises from the increasing demand for energy-efficient recommender systems that can provide accurate recommendations with low energy consumption and delay levels. To address this need, we propose a novel approach that combines the use of Fourier, Cosine, LSTM, GRU, Wavelet, and Gabor features for the representation of collected network datasetsand samples. These features are processed using low-power recurrent neural networks (LP RNN) to reduce energy consumption and delay levels. To further improve the accuracy and efficiency of the recommender engine, we apply an augmented post-evaluation Q-learning process. Our proposed method is compared with existing models, and we demonstrate its superior performance in terms of accuracy, precision, recall, energy consumption, and delay levels. Experimental results on multiple datasets and samples show that our proposed approach achieves an accuracy of 98.9%, precision of 97.5%, and recall of 98.3% while consuming less energy and incurring lower delay compared to existing models. Overall, our proposed multiparametric recommender engine provides an efficient and effective solution for low-power applications that require accurate recommendations for data patterns.