Scalable Learning of Collective Behavior

Tang, Lei; Wang, Xufei; Liu, Huan

doi:10.1109/tkde.2011.38

Cited by 38 publications

(40 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For instance, Tang & Liu [19] perform spectral clustering on the link structure to extract latent features, while Tang et al [22] use a scalable k-means clustering of the links to produce latent features that are certain to be sparse (e.g., having few non-zero values). In each case, the new features are then used as features for a link-unaware supervised classifier, where learning uses only the labeled nodes and their new latent features.…”

Section: B Latent Feature and Latent Link Methods For Lbcmentioning

confidence: 99%

A hidden treasure? Evaluating and extending latent methods for link-based classification

Fleming

McDowell

Markel

2014

Proceedings of the 2014 IEEE 15th International Conference on Information Reuse and Integration (IEEE IRI 2014)

View full text Add to dashboard Cite

Abstract-Many information tasks involve objects that are explicitly or implicitly connected in a network, such as webpages connected by hyperlinks or people linked by "friendships" in a social network. Research on link-based classification (LBC) has studied how to leverage these connections to improve classification accuracy. This research broadly falls into two groups. First, there are methods that use the original attributes and/or links of the network, via a link-aware supervised classifier or via a nonlearning method based on label propagation or random walks. Second, there are recent methods that first compute a set of latent features or links that summarize the network, then use a (hopefully simpler) supervised classifier or label propagation method. Some work has claimed that the latent methods can improve accuracy, but has not adequately compared with the best non-latent methods. In response, this paper provides the first substantial comparison between these two groups. We find that certain non-latent methods typically provide the best overall accuracy, but that latent methods can be competitive when a network is densely-labeled or when the attributes are not very informative. Moreover, we introduce two novel combinations of these methods that in some cases substantially increase accuracy.

show abstract

Section: B Latent Feature and Latent Link Methods For Lbcmentioning

confidence: 99%

A hidden treasure? Evaluating and extending latent methods for link-based classification

Fleming

McDowell

Markel

2014

Proceedings of the 2014 IEEE 15th International Conference on Information Reuse and Integration (IEEE IRI 2014)

View full text Add to dashboard Cite

show abstract

“…Computation indicates that within the same time window, forecasting achieves highly consistent results with K-means clustering. Also the forum topics are represented using graphs [14]. In this graph the node represent the topics and the edge represent the similarity or relationship between the topics IV.…”

Section: Iiirelated Workmentioning

confidence: 99%

Forum Reply Classification Using Clustering

Kalaiarasi¹,

Rooba²

2015

International Journal of Advanced Research in Computer and Comm

View full text Add to dashboard Cite

Due to richness of information in forums, researchers are increasingly interested in mining knowledge from forums. From this observation, the forum posts and replies are clustered and analyzed in order to improve the user knowledge in the field. To harvest knowledge from the forum the contents must be downloaded. Forum board or thread is usually divided into multiple pages which are linked by page flipping links. The forum sites contain different pages like entry pages, thread pages and page flipping.The forum mining have three phases: preprocessing, mining the data by applying various data mining strategies such as clustering and post processing. In preprocessing raw data is transformed into a usable format, mainly by parsing and cleaning. While preprocessing, the pages are downloaded as the html file and the files are invoked into parsing and assign attributes like forum id, forum title, thread count, post count. The parsing process is accomplished; data cleaning process is applied to the downloaded post sets and automatically remove noise data and irrelevant data. Clustering algorithm is applied for the preprocessed data to groups the forums into various clusters. The clustering is accomplished by using all topics and sub topics of the forum. The four dimensions of clustering are number of posts/topics, average sentiment values/topics, positive percentage of posts/topics and negative percentage of posts/topics. The posts/topics dimension are determined by number of replies for a post, the sentiment values of this topics are identified from user replies, it describe the user opinion, the positive and negative dimensions are determined from user replies, describe the user perception in the posts. The positive and negative dimensions are also used to identifying the user attitude and pros and cons of the specific topics are discussed in the particular forum. In the post processing stage numbers of clusters are obtained. The obtained final clusters are grouped based on the topics with similar sentiment values and user opinions. Based on the sentiment values, the positive and negative posts are clustered for each thread. Information seekers, decision makers can benefit from this clustering. It simplifies the decision making process.

show abstract

“…To deal with scalability edge centric clustering to extract sparse social dimension an innovation is stated [3].…”

Section: Related Workmentioning

confidence: 99%

“…Each edge is considered as one data instance and the associated modes are the respective features. Along with the above technique a cosine similarity method to identify similarities between different node is installed [1][2].…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Prediction of Collective Behavior in Live Social Media

Jadhav¹,

Mhetre²

2014

IJCA

View full text Add to dashboard Cite

Human interest is the precious thing in the present world. Currently, online social networks such as Facebook, Twitter, Google+, LinkedIn, and Foursquare have become extremely popular all over the world and play a significant role in people's daily lives. Social media has provided simple communication and propagation of data, normally, through commenting, sharing and publishing. Now a day's social networking site performs more important activity for businesses or social work. If the interest of the user is already known then it can be easy target for any business or social activity. So it is important to find out behavior or interest of the users.This work, intend to predict behavior of users in social media. Therefore our proposed system is going to show the behavior of the social networking user by extracting the online social networking data. Then construction of node graph and community graph to forms the grouping of similar behavior users. And then perform the clustering and classification for getting more accurate behavioral result of user.

show abstract

Scalable Learning of Collective Behavior

Cited by 38 publications

References 33 publications

A hidden treasure? Evaluating and extending latent methods for link-based classification

A hidden treasure? Evaluating and extending latent methods for link-based classification

Forum Reply Classification Using Clustering

Prediction of Collective Behavior in Live Social Media

Contact Info

Product

Resources

About