Voting mechanisms are widely adopted for evaluating the quality and credibility of user-generated content, such as online product reviews. For the reviews that do not receive sufficient votes, techniques and models are developed to automatically assess their helpfulness levels. Existing methods serving this purpose are mostly centered on feature analysis, ignoring the information conveyed in the frequencies and patterns of user votes. Consequently, the accuracy of helpfulness measurement is limited. Inspired by related findings from prediction theories and consumer behavior research, we propose a novel approach characterized by the technique of iterative Bayesian distribution estimation, aiming to more accurately measure the helpfulness levels of reviews used for training prediction models. Using synthetic data and a real-world data set involving 1.67 million reviews and 5.18 million votes from Amazon, a simulation experiment and a two-stage data experiment show that the proposed approach outperforms existing methods on accuracy measures. Moreover, an out-of-sample user study is conducted on Amazon Mechanical Turk. The results further illustrate the predictive power of the new approach. Practically, the research contributes to e-commerce by providing an enhanced method for exploiting the value of user-generated content. Academically, we contribute to the design science literature with a novel approach that may be adapted to a wide range of research topics, such as recommender systems and social media analytics.

With cross-disciplinary academic interests increasing and academic advising resources over capacity, the importance of exploring data-assisted methods to support student decision making has never been higher. We build on the findings and methodologies of a quickly developing literature around prediction and recommendation in higher education and develop a novel recurrent neural network-based recommendation system for suggesting courses to help students prepare for target courses of interest, personalized to their estimated prior knowledge background and zone of proximal development. We validate the model using tests of grade prediction and the ability to recover prerequisite relationships articulated by the university. In the third validation, we run the fully personalized recommendation for students the semester before taking a historically difficult course and observe differential overlap with our would-be suggestions. While not proof of causal effectiveness, these three evaluation perspectives on the performance of the goal-based model build confidence and bring us one step closer to deployment of this personalized course preparation affordance in the wild.

Complements and substitutes are two typical product relationships that deserve consideration in online product recommendation. One of the key objectives of recommender systems is to promote cross-selling, which heavily relies on recommending the appropriate type of products in specific scenarios. Research on consumer behavior has shown that consumers usually prefer substitutes in the browsing stage whereas complements in the purchasing stage. Thus, it is of great importance to identify the complementary and substitutable relationships between products. In this article, we design a neural network based framework that integrates the textual content and non-textual information of online reviews to mine product relationships. For the textual content, we utilize methods such as LDA topic modeling to represent products in a succinct form called “embedding.” To capture the semantics of complementary and substitutable relationships, we design a modeling process that transfers the product embeddings into semantic features and incorporates additional non-textual factors of product reviews. Extensive experiments are conducted to verify the effectiveness of the proposed product relationship mining model. The advantages and robustness of our model are discussed from various perspectives.

Recent years have witnessed a rapid increase in online data volume and the growing challenge of information overload for web use and applications. Thus, information diversity is of great importance to both information service providers and users of search services. Based on a diversity evaluation measure (namely, information coverage), a heuristic method—FastCovC+S-Select—with corresponding algorithms is designed on the greedy submodular idea. First, we devise the CovC+S-Select algorithm, which possesses the characteristic of asymptotic optimality, to optimize information coverage using a strategy in the spirit of simulated annealing. To accelerate the efficiency of CovC+S-Select, its fast approximation (i.e., FastCovC+S-Select) is then developed through a heuristic strategy to downsize the solution space with the properties of information coverage. Furthermore, ample experiments have been conducted to show the effectiveness, efficiency, and parameter robustness of the proposed method, along with comparative analyses revealing the performance’s advantages over other related methods. The online appendix is available at https://doi.org/10.1287/ijoc.2017.0753 .

Behavioral attention to stock markets in open source web search has attracted remarkable effort for financial analytics and applications in recent years. Differently from existing explorations at the overall market level, this paper investigates associative patterns that link search volume changes and stock price movements at individual stock level, so as to be able to support the design of flexible trading strategies. In doing so, a method (namely EATAPSP) is introduced as an extension to temporal associative pattern mining for time-series data, which can effectively discover the association between search volumes and stock prices in a temporal logic manner with the “after” predicate. Furthermore, the real world data on all China's A-share stocks from Shanghai Stock Exchange is used to test the method, revealing the effectiveness of the method. Moreover, with the discovered patterns, a new stock trading strategy (i.e., EATAPSP-Trading) is designed, which obtains superior cumulative returns and significantly outperforms other strategies.

Chinese hacker communities are of interest to cybersecurity researchers and investigators. When examining Chinese hacker communities, researchers and investigators face many challenges, including understanding the Chinese language, detecting variations in topic evolution, and identifying key hackers with their specialty areas. Therefore, we are motivated to develop a framework for analyzing key hackers and emerging threats in Chinese hacker communities. Specifically, we develop a set of topic models for extracting popular topics, tracking topic evolution, and identifying key hackers with their specialty topics. We applied our framework to 19 major Chinese hacker communities. As a result, we identified five major popular topics, including trading, fraud prevention & identification, calling for cooperation, casual chat, and monetizing. Moreover, we found several trends related to new communication channels, new stolen cards of interest, and new operating mechanism. Further, we also found the key hackers in each extracted area. Our work contributes to the cybersecurity literature by providing an advanced and scalable framework for analyzing Chinese hacker communities.

The consistency between review summaries and review ranking lists is important for consumers so they can utilize online reviews effectively and efficiently in their purchase decisions. This paper examines this consistency issue and formulates it as an optimization problem. Based on consumers’ reading behaviors, all possible sets of reviews that consumers would read from ranking lists are considered; the objective is to maximize the expected consistency. Because of the NP-hardness of the problem, exact methods that search for the optimal ranking lists are generally not acceptable in practice. Hence, a heuristic approach (the enhanced stepwise optimization procedure) is proposed. This approach is an effective and efficient approximation that selects reviews iteratively to add to the ranking lists in light of expected consistency value, superiority, and execution time. Intensive experiments on both synthetic and real data are conducted, with various environments and settings, along with a relevant user study, revealing that the proposed approach outperforms other related methods.

Competitiveness degree analysis is a focal point of business strategy and competitive intelligence, aimed to help managers closely monitor to what extent their rivals are competing with them. This article proposes a novel method, namely BCQ, to measure the competitiveness degree between peers from query logs as an important form of user generated contents, which reflects the “wisdom of crowds” from the search engine users’ perspective. In doing so, a bipartite graph model is developed to capture the competitive relationships through conjoint attributes hidden in query logs, where the notion of competitiveness degree for entity pairs is introduced, and then used to identify the competitive paths mapped in the bipartite graph. Subsequently, extensive experiments are conducted to demonstrate the effectiveness of BCQ to quantify the competitiveness degrees. Experimental results reveal that BCQ can well support competitors ranking, which is helpful for devising competitive strategies and pursuing market performance. In addition, efficiency experiments on synthetic data show a good scalability of BCQ on large scale of query logs.

Recent years have witnessed a rise of social publishing, which is a new type of social networking service. Social publishing has certain new features that call for a new way of managing and providing a large volume of documents. A fine data model is expected to evolve dynamically with the up-to-date knowledge, especially the associations that emerge in the context of social publishing. This paper first presents a conceptual schema of social publishing, which evolves to combine the association knowledge that reflects hidden associations in the data. A major constraint of concern is cardinality constraint. During the process of enriching a schema, the constraints to be specified should conform to the existing ones. A set of inference rules is presented for modeling with cardinality constraints. The rules are proven to be sound and complete, which helps to derive cardinality constraints from existing ones. The derived cardinality constraints are also proven to be consistent. © 2012 Wiley Periodicals, Inc.

Functional dependency (FD) is an important type of semantic knowledge reflecting integrity constraints in databases, and has nowadays attracted an increasing amount of research attention in data mining. Traditionally, FD is defined in the light of precise or complete data, and can hardly tolerate partial truth due to imprecise or incomplete data (such as noises, nulls, etc.) that may often exist in massive databases, or due to a very tiny insignificance of tuple differences in a huge volume of data. Based on the notion of functional dependencies with degrees of satisfaction (FDs)d, this article presents an efficient approach to discovering all satisfied (FDs)d using some important results obtained from exploration of (FDs)d properties such as extended Armstrong-like axioms and their derivatives. In this way, many dependencies can be inferred from previously discovered ones without scanning databases, and those unsatisfied ones could be filtered out inside (rather than after) the mining process. Fuzzy relation matrix operation is used to infer transitive dependencies in the mining algorithm. Finally, the efficiency is demonstrated with data experiments. © 2004 Wiley Periodicals, Inc. Int J Int Syst 19: 1089–1110, 2004.

Traditionally, the association rules discovered from categorical databases are like “Apples⇒Beers”, which means by default we focus on the attribute values which are equal to 1. Usually, we cannot deal with the association rules like “Age: 50–60 ∩ Female ⇒ ¬,Overdraw” whose semantics reflects “the Female users between 50 and 60 typically do not Overdraw ”. Here, we call this type of items (like “¬Overdraw”) as “opposite items”. In fact, however, in many categorical databases, “0” value does make sense. In this paper, we will propose a method to discover the association rules, which are composed of not only original items but also opposite items. Hereafter, Some optimizations are applied on the algorithm.