Research Topics List

CSC424/334 - Advanced Data Analysis

Following is a list of topics that you may want to consider for your in-class presentation and research paper. You are not limited to topics from this list. You may present on any topic directly related to Multivariate Data Analysis.

Remember that it is important to start thinking about a topic now. Additional details about schedule and format have been posted (see the Presentation/Research page). Remember that the first presentations will be scheduled on week 9. Also, you may work in groups of no more than two individuals (see last item on Presentation/Research page).
Note: Where possible, the papers below were either distributed in class and/or links to the papers in digital libraries have been provided. These libraries usually restrict access to subscribers but in some cases they may be accessed without charge through any workstation on the DePaul network including DPO subscribers.

Topics:

1. Multiple Regression:
   There are several possibilities. See the Software Engineering survey paper by Fenton that was distributed and discussed in class. This paper has several good references. This is also a topic where you may want to do an implementation. For example, you may implement an IML module to do model selection. For a model selection algorithm that you could try implementing, see the "Predicting Fault Detection Effectiveness" paper by Joseph Morgan et al., Proceedings of the Fourth International Software Metrics Symposium (1997).

2. Principal Component & Factor Analysis:
   There are several possibilities. See the papers handed out and discussed in class. This is a topic where you may want to analyze data that you are interested in studying. For example, you could try doing an analysis similar to that discussed in the "Somebody Wrote Shakespeare" paper by Ware Myers, IEEE Expert (1990). Also, this is a topic that lends itself to comparisons with other methods, a discussion of related methods, or an implementation using this method.
   • Your presentation could compare this method to Linear Discriminant Analysis.
   • You could discuss Factor Analysis. Remember that Factor Analysis is related to Principal Component Analysis. Since this was not covered in detail in class your presentation could provide an overview of this topic.
   • You could present the key features of an implementation such as that discussed in the "An Enhanced Neural Network Technique for Software Risk Analysis" paper by Donald Neumann, IEEE Transactions on Software Engineering (2002).
   Other papers of interest are:
   • "Multispectral Video", by P. Koponen et al., Proceedings of the International Conference on Pattern Recognition - ICPR'00 (2000)
   • "Human activity detection in MPEG sequences", by B. Ozer et al., Workshop on Human Motion - HUMO'00 (2000)
   • "Principal Components of Expressive Speech Animation", by Sumedha Kshirsagaret al., Proceedings of Computer Graphics Interational - CGI'01 (2001)
   • "AudiCom: A Video Analysis System for Auditing Commercial Broadcasts", by Juan Sanchez et al., IEEE International Conference on Multimedia Computing and Systems 2-II (1999)

3. Linear Discriminant Analysis:
   There are several possibilities. See the papers handed out and discussed in class. This is a topic where you may want to analyze data that you are interested in studying. Also, this is a topic that lends itself to comparisons with other methods or a discussion of related methods.
   • You could discuss Logistic Regression. Since this topic was not covered in class your presentation could provide an overview of this topic.
   • As mentioned above, your presentation could compare this method to Principal Component Analysis. See the "PCA versus LDA" paper by Alex Martinez et al., IEEE Transactions on Pattern Analysis and Machine Intelligence (2001).

4. Cluster Analysis:
   There are several possibilities. See the papers handed out and discussed in class for ideas. This is a topic where you may want to analyze data that you are interested in studying. Also, this is a topic that lends itself to comparisons of various clustering algorithms. For an interesting application of hierarchical clustering, see the "Interactively Exploring Hierarchical Clustering Results" paper by Jinwook Seo et al., IEEE Computer (2002). Also, see the "An Efficient k-Means Clustering Algorithm: Analysis & Implementation" paper by Tapas Kanungo et al., IEEE Transactions on Pattern Analysis & Machine Intelligence (2002).

• The IEEE Transactions on Pattern Analysis & Machine Intelligence is an excellent source for papers on Principal Component Analysis, Linear Discriminant Analysis, and related methods.
• The American Statistician is an excellent source of very readable papers on Multivariate Data Analysis.
• The IEEE Transactions on Software Engineering often contains papers that apply Multivariate Data Analysis to Software Engineering.
• The IEEE Computer magazine is a very readable source of papers on Multivariate Data Analysis.