Traffic Flow Forecast using Time Series Analysis based on Machine Learning

Intelligent Transportation System’s (ITS) main aim is to provide advanced services in both the transportation and traffic fields. A wide variety of algorithms and different types of models are being used for the estimation of short-term traffic flow. These algorithms works based on time series prediction and machine learning techniques for achieving improved results. Most of these models, on the other hand, require the historical data as continuous input, thus making it difficult to automatically find the best time delays. Our proposed study recommends a model "Long Short-Term Memory Recurrent Neural Network (LSTM RNN)" that employs the memory block's based on three multiplicative units to dynamically select the best time delays. The Performance Measurement System (PeMS) dataset is used for building this model that is then compared to other models like Random Walk (RW) and Support Vector Machine (SVM), one layer Feed Forward Neural Network (FFNN) and Stacking Auto Encoder (SAE). Based on the results it is observed that the proposed model gives better predictions than the other models and the same is measured in terms of accuracy.