Course: Data warehousing and mining

Instructions This document provides some guidelines for writing your project proposal and then your term paper. Note that the project is a significant portion of your grade (40%), so you are expected to devote a reasonable amount of time to it and to the write-up.

Types of Projects There are two main types of term paper projects. You can do a research project, where you look at a research issue. This could be original research or a topic already looked at. You can also examining real-world data sets and an associated problem. This is an application-based project. Ideally you should try to do something a bit interesting. You should make sure that your analysis is not trivial. For example, running a data set through WEKA and spending an hour on the analysis and then doing a quick write-up would be considered trivial.

Project Proposal (10%) Your project proposal must be typed and should be approximately 1 page long, single spaced. The purpose of the proposal is to make sure that you are on the right track and to give me enough information so that I can give you useful feedback. In your proposal you should cover the following items: Preliminary title and student information Abstract: Similar to the abstract that will ultimately appear in your paper. It should be one paragraph long, for now perhaps only 5-15 lines. It should provide a high-level summary of your project and outline your main goals. Brief description of what you plan to do.

o What problem are you trying to solve? A description of the problem area you wish to investigate. o How do you formulate the problem as a data mining problem (e.g., is it classification, association rule mining, etc.)? What exactly are you trying to predict (for prediction tasks) and how will you evaluate your results. How will you know if your results are good? What can you compare them to? It is critical that your problem is well-defined. o What data sets do you plan to use? If you must do significant work to get the data or convert it into the proper format, then describe the process and approximate effort required. How many examples are in the data set? Ho many features? o What learning tools do you plan to use (e.g., WEKA. Python Scikit) and what algorithms do you plan to use (e.g., decision trees, neural networks, etc.)? o List a few related research papers.

Project Write-up (20%) Your project paper or report is the main deliverable and should be about three (3) pages single-spaced. Project reports or papers that are too short will be deducted points. As a guide, your project report should contain the following sections and address the following questions.

Abstract: summarizes the paper and the goals of the work. It should be limited to a single paragraph. It should not provide a comprehensive summary of the paper. Rather it should motivate the problem, define it, and briefly discuss the general approach.

Introduction: Introduces the project and what you are trying to do. Should motivate the problem, quickly define it and the approach taken, and may discuss some highly related work. Probably should mention and contributions of the work.

Related Work: A description of related work, with citations to relevant papers. If you are doing an application paper, where you analyse some data, you are less likely to rely on a lot of related work. Nonetheless, there almost always should be some related work discussed. Your paper should mention a minimum of 5-7 related work papers. Experiment Methodology: Describes the experiments and the experiment methodology. Will describe the data sets, evaluation metrics, data mining algorithms used, the precise methodology related to the setup of experiments, and any other details related to the experiments.

Results: Presents the experiment results and a discussion and analysis of the results. Normally a separate discussion section is not necessary. Make good use of Tables and Figures.

Conclusion: Provide your conclusion (perhaps summarize your main results).Normally will also discuss limitations and avenues for future work.

References: Each paper should have a references section. This should include references to related work. The paper need not be organized exactly as described above, but it should be quite similar, since the outline above is generally what is used for most conference papers in data mining related conferences.