Abstract-Traffic speed is a key indicator for the efficiency of an urban transportation system. Accurate modeling of the spatiotemporally varying traffic speed thus plays a crucial role in urban planning and development. This paper addresses the problem of efficient fine-grained traffic speed prediction using big traffic data obtained from static sensors. Gaussian processes (GPs) have been previously used to model various traffic phenomena, including flow and speed. However, GPs do not scale with big traffic data due to their cubic time complexity. In this work, we address their efficiency issues by proposing local GPs to learn from and make predictions for correlated subsets of data. The main idea is to quickly group speed variables in both spatial and temporal dimensions into a finite number of clusters, so that future and unobserved traffic speed queries can be heuristically mapped to one of such clusters. A local GP corresponding to that cluster can then be trained on the fly to make predictions in real-time. We call this method localization. We use non-negative matrix factorization for localization and propose simple heuristics for cluster mapping. We additionally leverage on the expressiveness of GP kernel functions to model road network topology and incorporate side information. Extensive experiments using real-world traffic data collected in the two U.S. cities of Pittsburgh and Washington, D.C., show that our proposed local GPs significantly improve both runtime performances and prediction accuracies compared to the baseline global and local GPs.

The populations in many developed countries throughout the world are aging rapidly and the number of geriatric patients is expected to rise steeply in those countries. This will exert greater pressures on the management of hospital resources as a result. Hospital length of stay (LOS) is an important indicator of hospital activity and management because of its direct relation to resource consumption. Planning of hospital resources according to identified trends of LOS is, thus, an effective way to meet such future needs. In this paper, the authors propose a method to analyze the temporal trends of LOS based on the Coxian phase-type distributions, a special type of continuous-time Markov process. By fitting and regressing the probabilities of discharge from each phase of the distribution on time, the authors have found a growing trend in the proportion of long-staying patients in their sample of stroke patients from a general hospital in Singapore. The authors compare the yearly, quarterly and monthly trends over the same period to see the common pattern. The datasets were also robustified by bootstrapping to aid the analysis.

In this paper, we analyze the network of expertise constructed from the interactions of users on the online questionanswering (QA) community of Stack Overflow. This community was built with the intention of helping users with their programming tasks and, thus, questions are expected to be highly factual. This also indicates that the answers one provides may be highly indicative of one's level of expertise on the subject matter. Therefore, our main concern is how to model and characterize the user's expertise based on the constructed network and its centrality measures. We used the user's reputation established on Stack Overflow as a direct proxy to their expertise. We further made use of linear models and principal component analysis for the purpose. We found out that the current reputation system does a decent job at representing the user's expertise and that focus matters when answering factual questions. However, our model was not able to capture the other larger half of reputation which is specifically designed to reflect a user's trustworthiness besides their expertise. Along the way, we also discovered facts that have been known in earlier studies of the other/same QA communities such as the power-law degree distribution of the network and the generalized reciprocity pattern among its users.Recently, Anderson et al. [5] studied the Stack Overflow community with the purpose of characterizing and discovering long-lasting valued questions and answers in the community in order to promote their prominence and reduce the search effort. Their most relevant result is the proposed "reputation pyramid" model of answering behavior, i.e., when a question is posted, it is first attempted by the highly reputed users and

No abstract