Research Article
Hybrid Machine Learning Model for Traffic Forecasting
@INPROCEEDINGS{10.1007/978-3-030-76063-2_14, author={Khezaz Abderraouf and Manolo Dulva Hina and Hongyu Guan and Amar Ramdane-Cherif}, title={Hybrid Machine Learning Model for Traffic Forecasting}, proceedings={Science and Technologies for Smart Cities. 6th EAI International Conference, SmartCity360°, Virtual Event, December 2-4, 2020, Proceedings}, proceedings_a={SMARTCITY}, year={2021}, month={5}, keywords={Traffic forecasting Data augmentation Convolutional Neural Network K-Nearest Neighbour Deep Learning}, doi={10.1007/978-3-030-76063-2_14} }- Khezaz Abderraouf
Manolo Dulva Hina
Hongyu Guan
Amar Ramdane-Cherif
Year: 2021
Hybrid Machine Learning Model for Traffic Forecasting
SMARTCITY
Springer
DOI: 10.1007/978-3-030-76063-2_14
Abstract
Traffic prediction has been extensively studied in the past decades. Vehicle’s speed is considered the main factor for traffic forecasting, but external parameters, such as the weather, can also have a strong impact. This is a case of a classification problem to which Machine Learning has shown to have strong solving potential, if trained properly. In this paper, we propose a two-level model related to traffic forecasting parameters: It is necessary that there is no missing data in the training set, then train a Neural Network able to accurately predict the traffic situation Three completion algorithms from different types (Machine learning, algebraic and statistical methods) are compared for the rebuilding of the training set. The set is then used to train a Convolutional Neural Network into predicting the state of the traffic the way a human would do. The model is evaluated on the two parts: How accurately it can complete the data set and how correct the predictions are. This work is part of the ongoing research on intelligent vehicles that are capable of determining the context of the driving environment.
