Adaboost Triggered Extreme Learning Machine Algorithm for Traffic Flows Prediction in Vehicular Networks

Abstract

With the advent of vehicular IoT, smart transportation systems play a vital role in today's life demanding the prediction of the vehicular traffic flows and efficient data transfer.  At the same time, through machine learning (ML) algorithms, prediction of traffic flow has reached its new dimension but still usage of conventional machine learning algorithms needs improvisation in terms of prediction accuracy. Further so far, the literature work has not investigated for an Indian road scenario which serves as the motivation for this research study. Henceforth, this paper proposes a modified (ML) approach referred to as “AdaBoost Triggered Extreme Learning Machine algorithm” (ABT-ELM) for predicting high accurate traffic flows in Puducherry Union Territory, India. In this work, ELM predictor is trained using the training samples obtained through the Ada-boost algorithm. This algorithm is applied to obtain the ensemble weights of each layer of ELM predictor which is responsible for the improved prediction levels. Nearly two months of real-time traffic datasets from union territory of Puducherry, India are analysed and it as well serves as the test data-set for evaluating the proposed algorithm, using SUMO with OMNET++ platforms. The empirical study demonstrates that the proposed hybrid ABT-ELM learning approach outperforms classical learning approaches in terms of accuracy, precision and recall. Henceforth, it is suggested that the proposed ABT-ELM learning approach is highly promising and suitable for traffic prediction and management systems in Indian scenario.