A Study on the Role of Optimizers in Short-Term Traffic Flow Prediction Accuracy
Main Article Content
Abstract
This work demonstrates the creation of a strong and reliable model for predicting short-term traffic. The model utilizes Neural Prophet and is optimized using ADAM and (Stochastic Gradient Descent) SGD. Additionally, momentum is incorporated to improve the accuracy of predictions and the speed at which the model converges. Accurate anticipation of traffic congestion is crucial for effective urban planning and immediate traffic control. Conventional forecasting techniques frequently encounter difficulties in dealing with the ever-changing and intricate nature of traffic patterns. Neural Prophet is an advanced instrument for forecasting time-series data that use a complex method to accurately capture patterns in the data.This model gains advantages in efficiently managing large-scale data and reducing computational costs by including modern optimization techniques like ADAM and SGD. Incorporating momentum enhances the speed at which the convergence process occurs, ensuring stability and reducing oscillations while training.From the results we obtained the the prediction graph.Prediction graphs display data in many time intervals, including weekly, yearly, daily, and trends.