Spatiotemporal traffic forecasting plays a critical role in intelligent transportation systems, which empowers diverse urban services. Existing traffic forecasting frameworks usually devise various learning strategies to capture spatiotemporal correlations from the perspective of volume itself. However, we argue that previous traffic predictions are still unreliable due to two aspects. First, the influences of context factor-wise interactions on dynamic region-wise correlations are under exploitation. Second, the dynamics induce the credibility issue of forecasting that has not been well-explored. In this paper, we exploit the informative traffic-related context factors to jointly tackle the dynamic regional heterogeneity and explain the stochasticity, towards a credible uncertainty-aware traffic forecasting. Specifically, to internalize the dynamic contextual influences into learning process, we design a context-cross relational embedding to capture interactions between each context, and generate virtual graph topology to dynamically relate pairwise regions with context embedding. To quantify the prediction credibility, we attribute data-side aleatoric uncertainty to contexts and re-utilize them for aleatoric uncertainty quantification. Then we couple a dual-pipeline learning with the same objective to produce the discrepancy of model outputs and quantify model-side epistemic uncertainty. These two uncertainties are fed through a spatiotemporal network for extracting uncertainty evolution patterns. Finally, comprehensive experiments and model deployments have corroborated the credibility of our framework.

BACKGROUND: We recently developed the Paired End diTag (PET) strategy for efficient characterization of mammalian transcriptomes and genomes. The paired end nature of short PET sequences derived from long DNA fragments raised a new set of bioinformatics challenges, including how to extract PETs from raw sequence reads, and correctly yet efficiently map PETs to reference genome sequences. To accommodate and streamline data analysis of the large volume PET sequences generated from each PET experiment, an automated PET data process pipeline is desirable. RESULTS: We designed an integrated computation program package, PET-Tool, to automatically process PET sequences and map them to the genome sequences. The Tool was implemented as a web-based application composed of four modules: the Extractor module for PET extraction; the Examiner module for analytic evaluation of PET sequence quality; the Mapper module for locating PET sequences in the genome sequences; and the Project Manager module for data organization. The performance of PET-Tool was evaluated through the analyses of 2.7 million PET sequences. It was demonstrated that PET-Tool is accurate and efficient in extracting PET sequences and removing artifacts from large volume dataset. Using optimized mapping criteria, over 70% of quality PET sequences were mapped specifically to the genome sequences. With a 2.4 GHz LINUX machine, it takes approximately six hours to process one million PETs from extraction to mapping. CONCLUSION: The speed, accuracy, and comprehensiveness have proved that PET-Tool is an important and useful component in PET experiments, and can be extended to accommodate other related analyses of paired-end sequences. The Tool also provides user-friendly functions for data quality check and system for multi-layer data management.