Search

The detection of anomalies in spatiotemporal traffic data is not only critical for intelligent transportation systems and public safety but also very challenging. Anomalies in traffic data often exhibit complex forms in two aspects, (i) spatiotemporal complexity (i.e. we need to associate individual locations and time intervals formulating a panoramic view of an anomaly) and (ii) multi-source complexity (i.e. we need an algorithm that can model the anomaly degree of the multiple data sources of different densities, distributions and scales). To tackle these challenges, we proposed a three-step method that uses factor analysis to extract features, then uses the goodness-of-fit test to obtain the anomaly score of a single data point and then uses one class support vector machine to synthesize the anomaly score. Finally, we conduct extensive experiments on real-world trip data include taxi and bike data. And these extensive experiments demonstrate the effectiveness of our proposed approach. [ABSTRACT FROM AUTHOR]

Most existing heterogeneous information network (HIN) embedding methods focus on static environments while neglecting the evolving characteristic of real-world networks. Although several dynamic embedding methods have been proposed, they are merely designed for homogeneous networks and cannot be directly applied in heterogeneous environments. To tackle above challenges, we propose a novel framework for incorporating temporal information into HIN embedding, named multi-view dynamic HIN embedding (MDHNE), which can efficiently preserve evolution patterns of implicit relationships from different views in updating node vectors over time. We first transform HIN to a series of homogeneous networks corresponding to different views. Then our proposed MDHNE applies recurrent neural network (RNN) to incorporate evolving pattern of complex network structure and semantic relationships between nodes into latent embedding spaces, and thus the node vectors from multiple views can be learned and updated when HIN evolves over time. Moreover, we come up with an attention-based fusion mechanism, which can automatically infer weights of latent vectors corresponding to different views by minimizing the objective function specific for different mining tasks. Extensive experiments clearly demonstrate that our MDHNE model outperforms state-of-the-art baselines on three real-world dynamic datasets for different network mining tasks. [ABSTRACT FROM AUTHOR]

This paper addresses the anomaly detection problem in feature-evolving systems such as server machines, cyber security, financial markets and so forth where in every millisecond, |$N$| -dimensional feature-evolving heterogeneous time series are generated. However, due to stochasticity and uncertainty in evolving heterogeneous time series coupled with temporal dependencies, their anomaly detection are extremely challenging. Furthermore, it is practically impossible to train an anomaly detection model per single time series across millions of metrics, leave alone memory space required to maintain the model and evolving data points in memory for timely processing in feature-evolving data streams. Thus, this paper proposes |$\underline{o}$| ne sketch |$\underline{f}$| its all |$\underline{a}$| lgorithm (OFA), which is a real-time stochastic recurrent deep neural network anomaly detector built on assumption-free probabilistic conditional quantile regression with well-calibrated predictive uncertainty estimates. The proposed framework is capable of detecting anomalies robustly, accurately and efficiently in real time while handling randomness and variabilities in feature-evolving heterogeneous time series. Extensive experiments and rigorous evaluation on large-scale real-world data sets showcase that OFA outperforms other competitive state-of-the-art anomaly detector methods. [ABSTRACT FROM AUTHOR]

Finding groups in networks is very common in many practical applications, and most work mainly focus on dense groups. However, in scenarios like reviewer selection or weak social friends recommendation, we need to emphasize the privacy of individuals or minimize the possibility of information dissemination. So the internal relationship between individuals should be as tenuous as possible, but existing works cannot suit well to the requirement. Some works have focused on finding tenuous groups. However, these works only aim to find the most tenuous group and do not consider containing certain vertices. In this paper, we study the problem of finding tenuous groups in attributed networks that contain specific vertices. We first propose a new problem called Tenuous Attributed Group Query, and a new indicator, k-tenuity, to measure the structural tenuity of a group. Then we propose a method TAG-Basic to find proper groups by gradually selecting the vertices with optimal influence. We further design an advanced method TAG-ADV to improve the efficiency by forming a candidate set before selecting the optimal vertex. Experiment results show that k-tenuity is more effective than other state-of-the-art measurements, and our methods obtain the best result on group quality compared with other benchmark methods. [ABSTRACT FROM AUTHOR]

Predicting the bike demand can help rebalance the bikes and improve the service quality of a bike-sharing system. A lot of works focus on predicting the bike demand for all the stations, which is unnecessary as the travel cost of rebalance operations increases sharply as the number of stations increases. In this paper, we propose a framework for predicting the hourly bike demand based on the central stations we define. Firstly, we propose Two-Stage Station Clustering Algorithm to assign central stations and common stations into each cluster. Secondly, we propose a hierarchical prediction model to predict the hourly bike demand for every cluster and each central station progressively. Thirdly, we use a well-studied queuing model to determine the target initial inventory for each central station. The most innovative contribution of this paper is proposing the concept of central station, the use of a novel algorithm to cluster the central stations and present a hierarchical model, containing the Time and Weather Similarity Weighted K-Nearest Neighbor Algorithm and a linear model to predict the bike demand for central stations. The experimental results on the New York citi bike system demonstrate that our proposed method is more accurate than other methods in solving existing problems. [ABSTRACT FROM AUTHOR]

Due to the positive impact of ride sharing on urban traffic and environment, it has attracted a lot of research attention recently. However, most existing researches focused on the profit maximization or the itinerary minimization of drivers, only rare work has covered on adjustable price function and matching algorithm for the batch requests. In this paper, we propose a request matching algorithm and an adjustable price function that benefits drivers as well as passengers. Our request-matching algorithm consists of an exact search algorithm and a group search algorithm. The exact search algorithm consists of three steps. The first step is to prune some invalid groups according to the total number of passengers and the capacity of vehicles. The second step is to filter out all candidate groups according to the compatibility of requests in same group. The third step is to obtain the most profitable group by the adjustable price function, and recommend the most profitable group to drivers. In order to enhance the efficiency of the exact search algorithm, we further design an improved group search algorithm based on the idea of original simulated annealing. Extensive experimental results show that our method can improve the income of drivers, and reduce the expense of passengers. Meanwhile, ride sharing can also keep the utilization rate of seats 80%, driving distance is reduced by 30%. [ABSTRACT FROM AUTHOR]

Community detection plays a significant role in understanding the essence of a network. A recently proposed algorithm Attractor, which is based on distance dynamics, can spot communities effectively, but it depends on a cohesion parameter. Moreover, no efficient way is provided to find an optimal cohesion parameter setting. In this paper, we propose a parameter-free community detection algorithm by synchronizing distances iteratively. In each iteration, the distance of each edge will change dynamically according to the effect generated by its related neighbours. Several iterations later, distances between vertices belonging to the same community will synchronize to 0, while distances between vertices not in the same community will synchronize to 1. Besides, merging and division strategies are built up in the process of community detection. Experiments on both real-world and synthetic networks demonstrate benefits of our method compared to the baseline methods. [ABSTRACT FROM AUTHOR]