To alleviate or eliminate traffic congestion, the importance of soft measures for the effective use of existing roads is increasing. One such method involves providing drivers with future traffic condition predictions to encourage them to use roads during less congested times. To achieve this, AI traffic jam prediction technology was proposed, which focuses on the movement of people, the source of traffic demand, and predicts traffic conditions several hours to half a day in advance based on the population distribution of the day. However, the feasibility of AI traffic jam prediction has only been demonstrated on highways. Therefore, this study applied the AI traffic jam prediction technology to general roads and evaluated the prediction accuracy of travel time to confirm its feasibility. The evaluation results showed that the prediction accuracy was higher compared to baselines, such as historical statistical values and predictions based on multiple regression analysis, demonstrating the feasibility of AI traffic jam prediction on general roads.

This study proposes a non-parametric estimation method of the survival probability for road facilities using arbitrarily left-truncated, right-censored and interval censored data. Traditionaly, several studies have been proposed on the estimation of the survival probability especially in the medical and statistical fields. Although they deal with truncated and censored data, all of them assume that the left truncated time, a period from the survival start time to measurement start time, is known. However, for road facilities, once we miss the facility installation time, it is difficult to find the survival start time afterwards. Also, some studies assume the truncation and censoring timings are random. However, actual truncation and censoring timimings may not be perfectly random. Therefore, this study theoretically proposes an estimation method of the survival probability without knowing the truncated time using arbitrarily truncated, censored data; and validates the method using hypothetical data.