Hauptseminar "Mining Volatile Data" im WS 2012/13

News

• Final grades submitted on 7/12/2012.
• Feedback phase over.
• Presentations part 2 will take place on 14/11/2012. A short overview can be found here.
• Presentations part 1 took place on 7/11/2012. A short overview can be found here. The slides for each presentation will be uploaded soon.
• Check the detailed presentations program.
• For the paper assignment, send me an email (ntoutsi@dbs.ifi.lmu.de) with the following information: (i) your name (ii) your matrikelnummer, (iii) your favorite 3 papers in ranked order (e.g., 5, 3, 9).
• Kick-off meeting on Wednesday 19/09/2012, at 12:00 in Raum 161 (Oettingenstr. 67) slides
• The topics for the seminar are available for download.
• Sign up now.

Contents

Volatile data, that is, data that change over time, are met in many modern applications. A prominent example is social networks such as Twitter and Facebook which tend to be the new trend of our society. The content published in such a network evolves over time as member interests change over time either slightly (drift) or drastically (shift). The interactions among the members also change over time as new people enter/quit the network or the communication between the group members becomes more/ less intense.

Volatile data impose new challenges for data management due to their size and high degree of variability. Among these challenges are: efficient pattern extraction, adapting patterns to the underlying population changes, discovery of patterns that span more than one time point, change detection and monitoring.

In this seminar, we will study methods and algorithms for knowledge extraction and data mining in volatile data environments. An emphasis will be put on the area of social networks.

Organisation

• Contact person: Dr. Eirini Ntoutsi
• You have to sign up in UniWorx for the seminar.
• The kick off meeting will take place on September 19.
• The presentations will take place in November.

ECTS points:

• Participation is equivalent to 6 ECTS credits.
• The seminar is for both Master and Diploma students.

Requirements for the ECTS points: For receiving the ECTS-points, (i) a short summary (1 - 5 pages), (ii) a successful oral presentation as well as (iii) an active attendance during the whole seminar are required.

The oral presentation time is 25 minutes, followed by 10 minutes discussion. The short summary has to be provided to the lecturer till the day of the oral presentation.

Anmeldung

• You have to sign up in UniWorx for the seminar.

Papers

Below is a list of the papers, organized into different topics (already-assigned papers in gray color.):

Topic detection:
[1] Discovering evolutionary theme patterns from text: an exploration of temporal text mining
[2] Discovering Geographical Topics In The Twitter Stream
[3] Topic sentiment mixture: modeling facets and opinions in weblogs

Sentiment analysis:
[4] Sentiment Knowledge Discovery in Twitter Streaming Data

Community detection:
[5] A framework for community identification in dynamic social networks
[6] Microscopic evolution of social networks
[7] Finding and evaluating community structure in networks
[8] SCAN: A Structural Clustering Algorithm for Networks

Outliers & Events:
[9] Integrating Community Matching and Outlier Detection for Mining Evolutionary Community Outliers
[10] Event Detection in Social Streams
[11] On the Spatiotemporal Burstiness of Terms

Change detection & Evolution Monitoring:
[12] Temporal Structure Learning for Clustering Massive Data Streams in Real-Time
[13] Tracing Evolving Clusters by Subspace and Value Similarity

The papers for the seminar can be downloaded from within the Institute from here (rar, 6.62 MB).

Programm

• Wednesday, September 19, 2012, 12:00 kick-off meeting slides.
• Wednesday, November 7, 2012, Presentations part I intro slides.
• Wednesday, November 14, 2012, Presentations part II intro slides.

There are no other scheduled meetings for the seminar. On demand meetings can be arranged with each student for discussing his/her paper. Please send me an email (ntoutsi@dbs.ifi.lmu.de).

Place and time

Additional information

• How to give a lecture (pdf, 6K), by PD Dr. Matthias Schubert, in German.
• The DBLP Computer Science Bibliography
• Video lectures net
• KDnuggets