Mitigating Traffic Congestion Through Time Series Analysis and Deep Learning

Abstract

Action volume expectation is basic to asset assignment, blockage diminishment, and vitality utilization diminishment in activity overseeing frameworks. This investigation presents a novel approach utilizing time-series determining strategies combined with thick neural systems to effectively anticipate activity volume. Specifically, we see into five models: GRU, stacked LSTM, SARIMA, LSTM, and a combination of SARIMA and LSTM. We furthermore display a cross breed demonstrate that combines SARIMA and LSTM to upgrade forecast execution. Numerical measures like as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) are utilized to assess the models’ expectation precision. Our exploratory comes about illustrate that the recommended models perform uncommonly well in estimating activity volume, with the crossover show illustrating especially solid execution. Through this investigate, we trust to deliver solid and exact activity volume estimates that will move forward activity overseeing frameworks.