Unlike other public transportation services, the taxi service provides a convenient door-to-door way of transportation. It plays a uniquely important role in our daily life, especially when one travels to an unfamiliar place. The taxi service usually charges passengers based on the time or mileages they have taken. Taxi passengers can suffer from the risk of being overcharged on the unnecessary mileages incurred by taxi drivers intentionally or unintentionally. Due to the lack of background about the cities, most passengers cannot tell the subtle differences between the normal route and the altered one. The advances in location-acquisition and mobile computing techniques has generated massive taxi trajectory data from the GPS devices installed on taxis or the mobile devices carried by passengers. As the trajectory data contain rich temporal and spatial information of the behaviours of taxis, it is appealing to analyse such data to obtain informative insights. Moreover, anomaly analysis on taxi trajectory data can also contribute to traffic management which is one of the most important aspects of smart cities. This project aims to implement an anomalous taxi route detection system that can identify unusual routes by analysing the collected taxi trajectory data. The state-of-the-art machine learning algorithms will be leveraged to achieve the anomaly detection purpose. The main expected outcome is a mobile app together with cloud-based back-end servers. The data mining process on massive trajectory data will be conducted in cloud servers in order to achieve a scalable solution. Publicly available taxi trajectory datasets will be used to evaluate the implemented anomaly detection system. The students are expected to complete the following tasks: 1). Implementing the cloud-based architecture for the taxi route anomaly detection system; 2). Conducting experiments to compare the state-of-the-art anomaly detection methods on public available datasets, and selecting one or an ensemble of these methods for the project; 3). (Optionally) Proposing a novel trajectory anomaly detection algorithm to achieve better performance. Through this project, the students can expect to learn the following things: 1). Cloud-based programming paradigms, platforms and tools; 2). The state-of-the-art machine learning algorithms for trajectory data anomaly detection, and machine learning tools; 3). Geographical map related programming tools and APIs; 4). Research and development skills; 5). (Potentially) Academic publications.
Undergraduate
The main expected outcome is a mobile app together with cloud-based back-end servers.
Through this project, the students can expect to learn the following things: 1). Cloud-based programming paradigms, platforms and tools; 2). The state-of-the-art machine learning algorithms for trajectory data anomaly detection, and machine learning tools; 3). Geographical map related programming tools and APIs; 4). Research and development skills; 5). (Potentially) Academic publications.
None
Lab allocations have not been finalised