Nowadays, the rate of road incidents is continuously increasing as a result of elevated capability of vehicle acceleration that increases the risk of driver’s mistakes. Such road incidents directly impact the flow of traffic in such area and affect directly and indirectly to the economy, society, and environment. Incident monitoring and detection in Thailand is currently done by the responsible authority through CCTV and the traffic flow data from traffic flow measurement, both means of monitoring and detection have high operation costs. Online communication, on the other hand, has seen significant growth in the present days resulting in a fast growth of online social media use for various characteristic of communication replacing telephone calls. This article will present forms of incident detection from social media posts that have been data-mined from Twitter with autonomous API designed to screen for messages related to incident detection consisting of 4 steps. The experiment demonstrated the ability of the proposed method to detect incidents in Thai language with the accuracy of 85.80%, the detection rate (DR) of 78.83%, and false alarm rate (FAR) of 21.17%, based on the top 5 ranked keywords, out of 20 first keywords.
Keywords
Twitter, Incident Detection, Word Cloud
ECTI TRANSACTIONS ON COMPUTER INFORMATION TECHNOLOGY