The National Household Travel Survey (NHTS) is the largest survey in the transportation field. The main purpose of the survey is to build an O/D (origin/destination) matrix. O/D is a basic input in demand forecasting, urban planning, and so on.

The reduced data sample has becom an emerging issue in recent studies. For example, the sample rate was 2.41% in 2011, 1.13% in 2016, and 0.25% in 2021. A low sample causes a “zero-cell problem” when NHTS samples are expanded to the population.

Meanwhile, mobility, such as smartcard data and taxi trip record data, has been considered as a solution to this issue.

This research examines the existing mobility data and proposes Artificial Intellignece (AI) approach to improve O/D reliability. Multiple classification models and generative models are proposed using NHTS, smartcard data, taxi trip record data, mobile phone data, and land use data.

The model that estimated public transportation trip’s purpose and individual attributes showed a maximum of 86% accuracy. It is expected to improve the reliability and complexity of building not only the public transportation O/D but also private car OD. The model for taxis, however, is not usable since the sample size is too small. Collecting more samples a using simplified mobile survey is recommended to improve the AI model and also its usability.