Integrating Twitter Traffic Information with Kalman Filter Models for Public Transportation Vehicle Arrival Time Prediction

Citation

Abidin AF, Kolberg M & Hussain A (2015) Integrating Twitter Traffic Information with Kalman Filter Models for Public Transportation Vehicle Arrival Time Prediction. In: Trovati M, Hill R, Anjum A, Zhu S & Liu L (eds.) Big-Data Analytics and Cloud Computing: Theory, Algorithms and Applications. Cham, Switzerland: Springer, pp. 67-82. https://doi.org/10.1007/978-3-319-25313-8_5

Abstract
Accurate bus arrival time prediction is key for improving the attractiveness of public transport, as it helps users better manage their travel schedule. This paper proposes a model of bus arrival time prediction, which aims to improve arrival time accuracy. This model is intended to function as a pre-processing stage to handle real world input data in advance of further processing by a Kalman Filtering model; as such, the model is able to overcome the data processing limitations in existing models, and can improve accuracy of output information. The arrival time is predicted using a Kalman Filter (KF) model, by using information acquired from social network communication, especially Twitter. The KF Model predicts the arrival time by filtering the noise or disturbance during the journey. Twitter is one example of a Big Data source that offers a huge amount of unstructured data that can be analyzed and utilized for improving arrival time predictions. Twitter offers an API to retrieve live, real-time (road traffic) information, and offers semantic analysis of the retrieved twitter data. Twitter data, which has been processed, can be considered as a new input for route calculations and updates. This data will be fed into KF models for further processing to produce a new arrival time estimation.

Keywords
Kalman Filter; Application Programming Interface; Twitter User; Large Spike; Twitter Data