AI-predictive approaches in conjunction with DEA models to estimate the performance of wireless devices

This paper combined a Data Envelopment Analysis method with Machine learning (and their metaheuristic-improved versions), Deep Learning, and time series models to assess and estimate the performance of wireless devices. The results showed that among predictive models - Multi-Layer Perceptron, Support Vector Regression, Random Forest, Long Short-Term Memory, Autoregressive Integrated Moving Average - for estimating the performance of ۱۵۰۳ smartphones, the Multi-Layer Perceptron algorithm had the lowest Mean Squared Error (MSE) with a rate of ۸.۴۸۵۳E-۰۵ and was selected to estimate the efficiency of a new smartphone. A comparison of the curves fitted by the models to the actual efficiency scores showed that the Multi-Layer Perceptron and Long Short-Term Memory fitted curves closely matched the original curve. The Gray Wolf Optimizer was used to improve the performance of the Support Vector Regression and Random Forest algorithms that had low prediction accuracy.

کلیدواژه ها:

Performance evaluation ، Wireless devices ، Data Envelopment Analysis (DEA) ، Machine Learning (ML) ، Deep Learning(DL) ، Time series ، Metaheuristic method