Improving Neural Network Using Jaya Algorithm with Opposite Learning for Air Quality Prediction

Abstract

The Multi-Layer Perceptron Neural Network (MLP) is the commonly used Feedforward Neural Network (FNN) for tackling classification and prediction problems. The efficiency of MLP relies on the appropriate selection of its weights and biases. Usually, a gradient-based technique is used for tuning the selection of these parameters during the learning process. This technique suffers from its slow convergence and being stuck in local optima. Predicting urban air quality is vital to prevent urban air pollution and improve the life of residents. The air quality index (AQI) is a quantitative air quality tool. In this paper, an enhanced Jaya optimization algorithm is used to improve the MLP outcome (called EOL-Jaya-MLP). The opposite-learning method is used to improve the algorithm search space exploration. A three-year dataset from air quality monitoring stations is used in this study. The proposed technique is compared against the original Jaya and six machine learning techniques. Interestingly, the EOL-Jaya-MLP outperforms other techniques when predicting the AQI.