A Point of Interest (POI) refers to a specific location that people may find useful or interesting. While a large body of research has been focused on identifying and recommending POIs, there are few studies on characterizing the life cycle of POIs. Indeed, a comprehensive understanding of POI life cycle can be helpful for various tasks, such as urban planning, business site selection, and real estate evaluation. In this paper, we develop a framework, named POLIP, for characterizing the POI life cycle with multiple data sources. Specifically, to investigate the POI evolution process over time, we first formulate a serial classification problem to predict the life status of POIs. The prediction approach is designed to integrate two important perspectives: 1) the spatial-temporal dependencies associated with the prosperity of POIs, and 2) the human mobility dynamics hidden in the citywide taxicab data related to the POIs at multiple granularity levels. In addition, based on the predicted life statuses in successive time windows for a given POI, we design an algorithm to characterize its life cycle. Finally, we performed extensive experiments using large-scale and real-world datasets. The results demonstrate the feasibility in automatic characterizing POI life cycle and shed important light on future research directions.

The service usage analysis, aiming at identifying customers' messaging behaviors based on encrypted App traffic flows, has become a challenging and emergent task for service providers. Prior literature usually starts from segmenting a traffic sequence into single-usage subsequences, and then classify the subsequences into different usage types. However, they could suffer from inaccurate traffic segmentations and mixed-usage subsequences. To address this challenge, we exploit a multi-label multi-view learning strategy and develop an enhanced frame-work for in-App usage analytics. Specifically, we first devise an enhanced traffic segmentation method to reduce mixed-usage sub-sequences. Besides, we develop a multi-label multi-view logistic classification method, which comprises two alignments. The first alignment is to make use of the classification consistency between packet-length view and time-delay view of traffic subsequences and improve classification accuracy. The second alignment is to combine the classification of single-usage subsequence and the post-classification of mixed-usage subsequences into a unified multi-label logistic classification problem. Finally, we present extensive experiments with real-world datasets to demonstrate the effectiveness of our approach.

In online karaoke, the decision process in choosing a song is different from that in music radio, because users usually prefer songs that meet their vocal competence besides their tastes. Traditional music recommendation methods typically model users' personalized preference for songs in terms of content and style. However, this can be improved by considering the degree of matching the vocal competence (e.g. pitch, volume, and rhythm) of users to the vocal requirements of songs. To this end, in this paper, we develop a karaoke recommender system by incorporating vocal competence. Along this line, we propose a joint modeling method named CBNTF by exploiting the mutual enhancement between non-negative tensor factorization (NTF) and support vector machine (SVM). Specifically, we first extract vocal (i.e., pitch, volume, and rhythm) ratings of a user for a song from his/her singing records. Since these vocal ratings encode users' vocal competence from three aspects, we treat these vocal ratings as a tensor, exploit an NTF method, and learn the latent features of users' vocal metrics. These factorized features are simultaneously fed into an SVM classifier and then we use the trained classifier to predict the overall rating of a user with respect to a song. In addition, we propose an enhanced objective function to exploit the mutual enhancement between NTF and SVM, and devise an effective method to solve this objective as a coupled least-squares optimization problem via a maximum margin framework. With the estimated model, we compute the similarity between users and songs in terms of pitch, volume and rhythm and recommend songs to users. Finally, we conduct extensive experiments with real-world online karaoke data. The results demonstrate the effectiveness of our method.

With the booming popularity of online social networks like Twitter and Weibo, online user footprints are accumulating rapidly on the social web. Simultaneously, the question of how to leverage the large-scale user-generated social media data for personal credit scoring comes into the sight of both researchers and practitioners. It has also become a topic of great importance and growing interest in the P2P lending industry. However, compared with traditional financial data, heterogeneous social data presents both opportunities and challenges for personal credit scoring. In this article, we seek a deep understanding of how to learn users’ credit labels from social data in a comprehensive and efficient way. Particularly, we explore the social-data-based credit scoring problem under the micro-blogging setting for its open, simple, and real-time nature. To identify credit-related evidence hidden in social data, we choose to conduct an analytical and empirical study on a large-scale dataset from Weibo, the largest and most popular tweet-style website in China. Summarizing results from existing credit scoring literature, we first propose three social-data-based credit scoring principles as guidelines for in-depth exploration. In addition, we glean six credit-related insights arising from empirical observations of the testbed dataset. Based on the proposed principles and insights, we extract prediction features mainly from three categories of users’ social data, including demographics, tweets, and networks. To harness this broad range of features, we put forward a two-tier stacking and boosting enhanced ensemble learning framework. Quantitative investigation of the extracted features shows that online social media data does have good potential in discriminating good credit users from bad. Furthermore, we perform experiments on the real-world Weibo dataset consisting of more than 7.3 million tweets and 200,000 users whose credit labels are known through our third-party partner. Experimental results show that (i) our approach achieves a roughly 0.625 AUC value with all the proposed social features as input, and (ii) our learning algorithm can outperform traditional credit scoring methods by as much as 17% for social-data-based personal credit scoring.

College students with financial difficulties refer to those whose families can hardly afford their high tuition in universities, and should be supported by modern funding system. Indeed, students' economic plight negatively impact their mental health, academic performance, as well as their personal and social life. While funding students in financial hardship is widely accepted, there is limited understanding and research on effectively identification of the qualifying students. Traditional approaches relying on advisers' personal assessments are inefficient, and such subjective judgements may not reflect the truth. To this end, in this paper, we explore the data mining techniques for identifying students who are qualified for financial support. Specifically, we investigate students' complex behaviors on campus from multiple perspectives, and develop a learning framework, named Dis-HARD, by jointly incorporating the heterogeneous features to predict the portfolio of stipends a given student should be awarded. Our framework formalizes the above problem as a multi-label learning problem. Along this line, we first extract discriminative features from three perspectives: (i) smartcard usage behavior, (ii) internet usage behavior and (iii) trajectory on campus. Then, we develop a linear loss function with regularization to solve this multi-label classification problem. In addition, to effectively exploit the students' similarity and label dependency, we incorporate the graph Laplacian and composite l2,1-norm into the regularization of our model, and develop are-weighted algorithm to achieve effective optimization. Finally, experiments on real-world data demonstrate that our method consistently provides better performance compared to the existing state-of-the-art methods.

In this paper, we study the problem of personalized restaurant recommendations. Specifically, we develop a probabilistic factor analysis framework, named RMSQ-MF, which has the ability in exploiting multi-source information, such as the users' task, their friends' preferences, and human mobility patterns, for personalized restaurant recommendations. The rationale of this work is motivated by two observations. First, people's preferences can be affected by their friends. Second, human mobility patterns can reflect the popularity of restaurants to a certain degree. Finally, empirical studies on real-world data demonstrate that the proposed method outperforms benchmark methods with a significant margin.

With lots of applications emerging in World Wide Web, many interaction data from users are collected and exploited to discover user behavior or interest patterns. In this paper, we attempt to exploit a new interaction data, namely print logs, where each record is printing URLs selected by a user using a popular web printing tool. Users usually print web contents based on an intention (subtask or task). Apparently, mining common print tasks from print logs is able to capture users' intentions, which undoubtedly benefits many web applications, such as task oriented recommendation and behavior targeting. However, it is not an easy job to perform this due to the difficulty of URL topic representation and task formulation. To this end, we propose a general framework, named UPT (Users Print Tasks mining framework), for mining print tasks from print logs. Specifically, we attempt to leverage delicious (a social book marking web service) as an external thesaurus to expand the expression of each URL by selecting tags associated with the domain of each URL. Then, we construct a tag co-occurrence graph where similar tags can be clustered as subtasks. If we view each subtask as an item, then the print log is transformed to a transaction database, on which an efficient pattern mining algorithm is proposed to induce tasks. Finally, we evaluate the effectiveness of the proposed framework through experiments on a real print log.