Seminar: Advanced Topics in Data Mining
- Contact person: Prof. Dr. Eirini Ntoutsi
- Participation is equivalent to 3 ECTS credits (1 ECTS credit equals 30 hours of study).
- We will use StudIP for discussions/announcements/material/schedule.
- Kick-off meeting: Wednesday 17/04/2019
- C4.5 & fairness-aware DT induction
- Naive Bayes & tackling the poor assumptions
- Naive Bayes & model merging
- Naive Bayes & fairness-aware NB
- Adaboost & cost-based learning
- Adaboost & fairness-aware adaboost
- kNN & class imbalance
- kNN & high dimensionality
- SVM & fairness-aware learning
- SVM & cost-sensitive learning
- CART & Gradient tree boosting
- k-Means & stability
- k-Means & Stream k-Means
- EM & dealing with outliers
- DBSCAN & Density-based stream clustering
- Apriori & Redundancy reduction
- Prof. Dr. Eirini Ntoutsi
- MSc Vasileios Iosifidis
- MSc Tai Le Quy
- MSc Felipe Reis
- MSc Amir Abolfazli
About the seminar
Goal of the seminar is the independent research of a scientific topic based on a publication as well as a high quality presentation of the topic in both written (report) and spoken forms (presentation and Q&A sessions). This seminar is dedicated to the discussion of selected topics in data mining. Each semester the focus is on a different topic, for example, certain learning techniques (such as clustering, classification, …), evaluation methods, mining for different data types (such as timeseries, trajectories, text …), feature selection etc.SoSe19 focus: Top algorithms in Data Mining
In SoSe19, we will focus on top algorithms for data mining. In particular, we will review top algorithms as well as certain extensions for each of them.Organisation
Process
The students choose a topic from the list of provided topics. For each topic, two papers will be provided: a paper referring to the original algorithm and another one refering to some extension of the algorithm. Students are expected to carefully read the papers as well as related work (at least 2 papers) necessary to comprehend the topics (i.e., both the original algorithm and the extension).For each student there will be an advisor who guides the whole process and helps the student in case of difficulties and questions. The students are expected to write a report of their research topic, present their findings to the class and be able to pose and answer questions in the Q&A sessions.
The final grade depends on the presentation, report and overall participation and engagement.
Schedule
There will be a few group meetings and regular meetings with the advisors through the semester. It is obligatory for the students to participate in those meetings and actively engage in discussions and Q&A sessions.