Iterative Skill Optimization based Recurrent Network for Air Quality Forecasting
Periasamy Sannasi, Parthiban Subramanian, R Surendran
Abstract
This research introduces a novel air quality prediction model using the optimized deep learning model for enhancing the forecasting accuracy by solving the issue of over-fitting. For this, Iterative Skill Optimization based Recurrent Network (Iter_SkO-RecurNet) model is proposed. The Iter_SkO algorithm is used to optimize the algorithm's hyper-parameters in order to minimize the loss function in the developed model, which uses LSTM for air quality forecasting. The loss function optimization is carried out using the suggested Iter_SkO algorithm, which incorporates the iterative chaotic mapping into the Skill optimization process. The iterative chaos introduces the randomness in the algorithm and assists to acquire global best solution. The analysis of the proposed model based on R2, RMSE, and MAPE and acquired the values of 0.9775, 0.92714775, and 8.174332 respectively.