Making a Research Proposal

The entire proposal may be no more than three pages (double spaced, 12 point font, 1 inch margins). The references do not count toward your page limits. The project proposal should be written in R Markdown and knitted to PDF.

Your proposal should be well written, reproducible, and well-organized.

Your project proposal should include all of the following elements:

1. Title Page: Include the names of all team members.

2. Introduction: Provide as much information about the research problem and the data. Try to answer the following questions. Why is the research problem of significant interest? What are the main research questions? What are the challenges? You should include a brief literature review. For the lit review, you should find peer-review articles related to the topic. Use a publication search engine such as Google Scholar.

3. Initial Hypotheses: Before you look at the data, provide a list of detailed hypotheses. E.g. you might believe that a particular set of features can be used for the detection of fraud. Use scientific arguments to support your belief.

4. Data-drivin Hypotheses: What new hypotheses are you planning to develop as you explore the data? Provide some description of things that you find interesting when you were looking at the problem and the data.

5. Proposed work and discussion: Desribe what you are proposing to do and provide a brief discussion of how your results may be placed in the context of the literature.

6. References: A list of references cited in your report. Use a standard format for references (such as APA or MLA).

Read the data description. Use different statistical learning methods covered in the course for regression and/or classification. You can use linear, logistic, polynomial regression with proper variable selection, linear or quadratic discriminant analysis, K-nearest neighbor classifier, jackknife, bootstrap, ridge regression, lasso, principal components regression, partial least squares, splines, regression and classification trees, support vector machines, clustering, and related methods. Apply cross-validation techniques to find the optimal degree of flexibility – the best subset of predictors or the optimal tuning parameters. Evaluate prediction performance of competing methods. Illustrate results with appropriate plots and diagrams.