Luiz Henrique de Campos Merschmann

Detecting tip spam in location-based social networks

Location Based Social Networks (LBSNs) are attracting new users in exponential rates. LBSNs like Foursquare and Gowalla allow users to share their geographic location with friends, search for interesting places as well as posting tips about existing locations. By allowing users to comment on locations, LBSNs increasingly have to deal with a new wave of spammers, which aim at advertising unsolicited messages on tips and comments about locations. In this paper, we investigated the task of identifying tip spam on a popular Brazilian LBSN system, namely Apontador. Based on a labeled collection of tips provided by Apontador as well as crawled information about users and locations, we identified a number of attributes able to distinguish spam from non-spam tips. We leveraged our characterization study towards a spam detection mechanism. Using a classification technique, we were able to correctly identify 84% of spam tips and 91.8% of non-spam tips. Our results also highlight the importance that places and related user activity have for detecting tip spam on LBSNs.

Categorizing feature selection methods for multi-label classification

In many important application domains such as text categorization, biomolecular analysis, scene classification and medical diagnosis, examples are naturally associated with more than one class label, giving rise to multi-label classification problems. This fact has led, in recent years, to a substantial amount of research on feature selection methods that allow the identification of relevant and informative features for multi-label classification. However, the methods proposed for this task are scattered in the literature, with no common framework to describe them and to allow an objective comparison. Here, we revisit a categorization of existing multi-label classification methods and, as our main contribution, we provide a comprehensive survey and novel categorization of the feature selection techniques that have been created for the multi-label classification setting. We conclude this work with concrete suggestions for future research in multi-label feature selection which have been derived from our categorization and analysis.

Lazy attribute selection: Choosing attributes at classification time

Attribute selection is a data preprocessing step which aims at identifying relevant attributes for the target machine learning task --namely classification in this paper. In this paper, we propose a new attribute selection strategy --based on a lazy learning approach --which postpones the identification of relevant attributes until an instance is submitted for classification. Our strategy relies on the hypothesis that taking into account the attribute values of an instance to be classified may contribute to identifying the best attributes for the correct classification of that particular instance. Experimental results using the k-NN and Naive Bayes classifiers, over 40 different data sets from the UCI Machine Learning Repository and five large data sets from the NIPS 2003 feature selection challenge, show the effectiveness of delaying attribute selection to classification time. The proposed lazy technique in most cases improves the accuracy of classification, when compared with the analogous attribute selection approach performed as a data preprocessing step. We also propose a metric to estimate when a specific data set can benefit from the lazy attribute selection approach.

Pollution, bad-mouthing, and local marketing: The underground of location-based social networks

Abstract Location Based Social Networks (LBSNs) are new Web 2.0 systems that are attracting new users in exponential rates. LBSNs like Foursquare and Yelp allow users to share their geographic location with friends through smartphones equipped with GPS, search for interesting places as well as posting tips about existing locations. By allowing users to comment on locations, LBSNs increasingly have to deal with new forms of spammers, which aim at advertising unsolicited messages on tips about locations. Spammers may jeopardize the trust of users on the system, thus, compromising its success in promoting location-based social interactions. In spite of that, the available literature is very limited in providing a deep understanding of this problem. In this paper, we investigated the task of identifying different types of tip spam on a popular Brazilian LBSN system, namely Apontador. Based on a labeled collection of tips provided by Apontador as well as crawled information about users and locations, we identified three types of irregular tips, namely local marketing, pollution and, bad-mouthing. We leveraged our characterization study towards a classification approach able to differentiate these tips with high accuracy.

An Extended Local Hierarchical Classifier for Prediction of Protein and Gene Functions

Gene function prediction and protein function prediction are complex classification problems where the functional classes are structured according to a predefined hierarchy. To solve these problems, we propose an extended local hierarchical Naive Bayes classifier, where a binary classifier is built for each class in the hierarchy. The extension to conventional local approaches is that each classifier considers both the parent and child classes of the current class. We have evaluated the proposed approach on eight protein function and ten gene function hierarchical classification datasets. The proposed approach achieved somewhat better predictive accuracies than a global hierarchical Naive Bayes classifier.

A VNS algorithm for feature selection in hierarchical classification context

Abstract Feature selection, usually adopted as a preprocessing step for data mining, is used to select a subset of predictive features aiming to improve the performance of a predictive model. Despite of the benefits of feature selection for classification task, to the best of our knowledge, there is no work in the literature that addresses feature selection in conjunction with global hierarchical classifiers. Thus, in this paper, we fill this gap proposing a feature selection method based on Variable Neighborhood Search (VNS) metaheuristic for the hierarchical classification context. Computational experiments were carried out on five bioinformatics datasets to evaluate the effect of the proposed algorithm on classification performance when using a global hierarchical classifier. As result, we have obtained a classifier performance improvement for three datasets and a competitive result for a fourth dataset, which indicates the suitability of the proposed method for the hierarchical classification scenario.