1. Hybrid Feature Based Label Generation Approach for Prediction of Traffic Congestion in Smart Cities
- Author
-
Aamish Izhar, S. A. M. Rizvi, and S. M. K. Quadri
- Subjects
050210 logistics & transportation ,Computer science ,Cumulative distribution function ,010401 analytical chemistry ,05 social sciences ,Binary number ,computer.software_genre ,01 natural sciences ,0104 chemical sciences ,Data modeling ,Support vector machine ,Traffic congestion ,Smart city ,0502 economics and business ,Feature based ,Data mining ,computer - Abstract
Traffic congestion is one of the major challenges in smart cities due to rapid urbanization. In this paper, we study the problem of traffic congestion prediction in smart city transportation systems. While there are various methods to tackle this problem, most of them suffer from improper label generation. Therefore, motivated by such shortcomings, we propose an intuitive and logical solution where we consider several road-related factors for the effective prediction of traffic congestion. For the same, we consider two datasets from the CityPulse EU FP7 project and employed two well-known binary classifiers. Our results indicate that labels generated based on the hybridization of the average speed of vehicles and the number of vehicles plying on the road prove to be effective in properly discriminating congestion and non-congestion scenarios. Moreover, number of vehicles alone can produce significant discrimination as well. Furthermore, the cumulative distribution function (CDF) applied to the above-mentioned factors with respect to the labels generated also validates the effectiveness of our approach.
- Published
- 2020
- Full Text
- View/download PDF