Machine Learning for Traffic Flow Prediction and Management in Urban Civil Infrastructure

Abstract

One issue that all urban centres have in common is traffic congestion on the road networks. Comprehending the movement of traffic along road segments is imperative for generating workable remedies; yet, this is an expensive undertaking, particularly for developing nations. Based on historical traffic flow data, geometric data, and Google Distance Matrix API data, this study suggests a cost-effective method for a directional flow forecast model for metropolitan roads. The collected data was combined in time and space to serve as model estimation attributes. Devi- ating from classic probability estimates, a K- Nearest Neighbour regression method was utilised in the investigation. A test dataset was used to validate the model; the results indicated that the mean absolute error of prediction and the root mean square error were, respectively, 2.318 and 9.479. When the geometry of the road is known, lane flow can be estimated using the journey time and speed data obtained from the Google Distance matrix API.