Recently, the rapid growth in the number of customer reviews on e-commence platforms and in the amount of user-generated content has begun to have a profound impact on customer purchasing decisions. To counter the negative impact of social media marketing, some firms have begun hiring people to generate fake reviews which either promote their own products or damage their competitor's reputation. This study proposes a framework, which takes advantage of both supervised and unsupervised learning techniques, for the observation of behaviors among spammers. Then, based on the behavior of participants on web forums, the authors build up a post-reply network. The main focus is on the behavior-related features of the reviews, their propagation, and their popularity. The primary objective of this study is to build an effective online spammer detection model and the method detailed in this work can be used to improve the performance of spammer detection models. An experiment is carried out with a real dataset, the results of which indicate that these new features are important for identifying spammers. Finally, random walk clustering is applied to investigate the post-reply network. Some interesting and important features are observed in the interactions between a group of spammers which could be subjected to further research.

In this study, we sought to elucidate the reading behaviors of users on an online domestic news platform. Our specific objective was to identify which features of online news positively impact the reading behaviors of users (i.e. click-through rate and news article popularity.) For this, we cooperated with Chinatimes.com, one of the largest news media companies in Taiwan. Two experiments were performed using data from the website analytics of Chinatimes.com and from Facebook's graph API: multiple linear regression and decision tree modeling. The results show that multiple linear regression is a significant predictor of news article popularity (R-squared = 0.7731) but not click-through rate. The decision tree model further revealed that users are more likely to read entertainment news articles when articles feature a greater number of images per page; political news is viewed more when the number of keywords is increased; travel, sports, and world news achieve article popularity through sharing on Facebook; and business news attracts more readers when the news source is Chinatimes.com.

Facebook consists of more than 80% of the individuals using instant messaging and social networking in the West. The Facebook fan page is an efficient way to interact with fans in order to enhance customer loyalty (fan engagement). This study analyzed regional ticket purchase behavior associated with marketing campaigns on Facebook with a focus on fan pages. We found that Facebook fan pages differ regionally with regard to their effect on the exhibition market in Taiwan. These findings provide a valuable resource in the accurate and marketing of events using social media in Taiwan.

In the big data era, data mining techniques and applications are becoming increasingly important in various industries. Among numerous data mining techniques, frequent patterns are a crucial tool. The majority of existing studies on frequent pattern mining used single minimum support thresholds, which is unreasonable in the real world. Although there have been a lot of extensive research on support thresholds, particularly multiple minimum supports, their mining processes have all been extremely time-consuming. With advances in hardware and the maturing of cloud computing and distributed computing, the costs of storage media are lower, so we can get a look at the entirety of data. The internet offers numerous conveniences, which produced today's highly competitive industry. Obtaining crucial business information quickly from big data so as to make business decisions is a company's big challenge and key to success. We therefore proposed a high efficiency frequent pattern mining algorithm with multiple minimum supports based on the CFP-growth method, which is a frequent pattern mining approach aimed at multiple minimum support environments. It is hoped that this approach can provide a more efficient and advanced mining method to obtain hidden information that is crucial for corporations to make swift and accurate business decisions with. Our experiment results indicated that in the same execution environment, the overall efficiency of the proposed method is superior to the CFP-growth method.

Structured continuous-label classification is a variety of classification in which the label is continuous in the data, but the goal is to classify data into classes that are a set of predefined ranges and can be organized in a hierarchy. In the hierarchy, the ranges at the lower levels are more specific and inherently more difficult to predict, whereas the ranges at the upper levels are less specific and inherently easier to predict. Therefore, both prediction specificity and prediction accuracy must be considered when building a decision tree (DT) from this kind of data. This paper proposes a novel classification algorithm for learning DT classifiers from data with structured continuous labels. This approach considers the distribution of labels throughout the hierarchical structure during the construction of trees without requiring discretization in the preprocessing stage. We compared the results of the proposed method with those of the C4.5 algorithm using eight real data sets. The empirical results indicate that the proposed method outperforms the C4.5 algorithm with regard to prediction accuracy, prediction specificity, and computational complexity.

In traditional decision (classification) tree algorithms, the label is assumed to be a categorical (class) variable. When the label is a continuous variable in the data, two possible approaches based on existing decision tree algorithms can be used to handle the situations. The first uses a data discretization method in the preprocessing stage to convert the continuous label into a class label defined by a finite set of nonoverlapping intervals and then applies a decision tree algorithm. The second simply applies a regression tree algorithm, using the continuous label directly. These approaches have their own drawbacks. We propose an algorithm that dynamically discretizes the continuous label at each node during the tree induction process. Extensive experiments show that the proposed method outperforms the preprocessing approach, the regression tree approach, and several nontree-based algorithms.