Developing an optimised activity type annotation method based on classification accuracy and entropy indices

The generation of substantial amounts of travel- and mobility-related data has spawned the emergence of the era of big data. However, this data generally lacks activity-travel information such as trip purpose. This deficiency led to the development of trip purpose inference (activity type imputation/annotation) techniques, of which the performance depends on the available input data and the (number of) activity type classes to infer. Aggregating activity types strongly increases the inference accuracy and is usually left to the discretion of the researcher. As this is open for interpretation, it undermines the reported inference accuracy. This study developed an optimised classification methodology by identifying classes of activity types with an optimal balance between improving model accuracy, and preserving activity information from the original data set. A sensitivity analysis was performed. Additionally, several machine learning algorithms are experimented with. The proposed method may be applied to any study area. This work was supported by the National Research Foundation of Korea funded by the Korean government (MSIP) under grant NRF-2010-0028693.