CMSC 828G
Principles of Data Mining
Spring 2002
Project

Overview:

A major component of the course is a research project. The goal of this project is to give you hands-on experience with the full range of tasks involved in data mining, including:

• choosing a domain
• creating a data set
• identifying analysis objectives
• choosing algorithms
• evaluating algorithms
• reporting conclusions

Ideally, your project should be of the quality that it can either be published or that someone would be willing to pay consulting fees for the results of your work.

There are seven project deliverables (i.e., things you will need to turn in and/or present):.

Detailed instructions for each project component will be handed out sometime before the due dates. Below are some high level instructions for each part.

P1. Data set selection

Choose a domain that you will explore in your project. You may use an existing data base or construct your own. Ideally, the data set will be about something you care about or are interested in (stock performance, sports statistics, music recommendations, etc.). Please do not use any of the standard testbeds such as data sets from the UCI Machine Learning Repository for your main data set.

The larger the data set, the better. At a minimum your dataset should have 2000 records, each with at least 7 attributes.

You will be required to turn in a one page description of your data set and give a brief (2-3 minute) description of your data set to the class.

An optional component is to make your data set available to others via the class web page. You will not get extra points for this (you may get kharma points, but no class points).

P2. Project Proposal

Decide on the data mining objective for your project, the algorithms you will use and your evaluation criteria.

examples:

• Your objective may be to learn a descriptive model of the domain, you may choose to contrast several clustering methods and you may evaluate the effectiveness using cross validation.
• Your objective may be to learn a predictive model, you may choose to contrast learning a decision tree with learning a support vector machine and you may evaluate the construction times for each of the methods for different data organizations.
• Your objective may be to learn a causal model, you may choose to contrast using a Bayesian network learning algorithm with using a constraint-based approach, and you may evaluate the accuracy of the learned causal models.

You should contrast at least two methods, or contrast several significant variations in an algorithm class. You may write your own code or you may use existing code. Pointers to existing implementations will be made available.

Another possibility is to choose a paper from one of the recent conferences and try to duplicate their results (and potentially extend them as well).

In your proposal, you should describe the objectives, algorithms (including whether you will use existing implementations or your own) and the evaluation criteria. Please also provide milestones for each week until the project is due.

P3. Project Progress Report

A one page description of what you have done so far on the project, and what is left to be done. Ideally you will be half way through.

P4. Project Report

The first draft of your project report. The report should be written as if it is being submitted for publication. It should have an abstract and an introduction that describes your objects, methods and evaluation. You should review related work, present your results and discuss your conclusions.

The following guidelines from Ray Mooney may also be helpful html, slides

You should turn in three copies of your project report.

P5. Project Review

Each person will be given two project reports to review. A review form will be provided. Reviews should give constructive feedback.

P6. Project Presentation

Each student will give an in class presentation of their project.

P7. Final Project Report

The final project report should address issues raised by the reviews.

Some important guidelines to follow:

• Do not reuse a project from another course you have previously taken or are currently taking. If you do intend to do a project similar to one that you have done in another class, or a project that will be used for another class in addition to cs828G, please be sure to discuss it with me and to clearly state this on your project proposal. Failure to do so will result in a 0 on the project.
• If you use code off the web, which is encouraged, please cite the source. You should cite the source even if you make modifications to the code.
• In your write-up, if you copy text verbatim from the web or other sources, you must not only reference the source but you should also have the passage in quotes in your write-up. Otherwise this is plagiarism.

CMSC 828G HOME

Web Accessibility