In the public transportation domain, accurate estimation of travel times helps to manage rider expectations as well as to provide a powerful tool for transportation agencies to coordinate the public transport vehicles. Although many statistical and machine learning methods have been proposed to estimate travel times, none of the methods consider utilizing predicted traffic information. Forecasting how congestion is going to evolve is critical for accurate travel time estimations. In this paper, we present DeepTRANS, which incorporates traffic forecasting information to our prior Deep Learning-based Bus Estimated Time of Arrival (ETA) model, increasing its accuracy by 21% in estimating bus travel time.

Real-time outlier detection in data streams has drawn much attention recently as many applications need to be able to detect abnormal behaviors as soon as they occur. The arrival and departure of streaming data on edge devices impose new challenges to process the data quickly in real-time due to memory and CPU limitations of these devices. Existing methods are slow and not memory efficient as they mostly focus on quick detection of inliers and pay less attention to expediting neighbor searches for outlier candidates. In this study, we propose a new algorithm, CPOD, to improve the efficiency of outlier detections while reducing its memory requirements. CPOD uses a unique data structure called "core point" with multi-distance indexing to both quickly identify inliers and reduce neighbor search spaces for outlier candidates. We show that with six real-world and one synthetic dataset, CPOD is, on average, 10, 19, and 73 times faster than M_MCOD, NETS, and MCOD, respectively, while consuming low memory.