Homework 10. Stat 540 Spring 1999.

Deliverables (by 4/9/99) :

• A. Code in R to perform the boosting algorithm.
• B.Code in Perl to extract the price of a stock from a Yahoo stock page , and a subroutine to convert it into a price.
• C.Code in Perl to extract all links from a web page, annotate your regexp to explain what it is doing ( the x modifier) as in the example on page 71 of the Perl book.
• D.Code in Perl to extract all citations from a Patent database page.

A. Use the dataset uvatoday.txt to perform the boosting. For this homework you do not have to do out of sample prediction, the didactic objective is to code the boosting algorithm. Use a multiple logistic regression (without interactions) as the weak learner. Notice how the mis-classification probability increases to 0.5 from one iteration to the next, as well as the corresponding log odds approaches 0. R commands that may be useful are:

 glm predict ifelse sum abs log

B. Use the LWP module to obtain a stock price WEB page. You will have to find two identifiers that bracket the stock price, and then go to work on extracting and manipulating what lies in between them. A good way to find 2 identifiers is to obtain a page using Netscape, then to use the VIEW SOURCE function to find the identifier.

C. You can probably find examples of this on the web, but if you do, make sure that you know what they are doing.

D. The above link takes you into the Patent database, but you can get an individual page by using a URL of the form http://patents.uspto.gov/cgi-bin/ilink4?INDEX+0+5615225+F. It's the last number that is the citation number. Once you have obtained one of these records, parse it, pulling off all the patents it cites. You code should produce a list ofpatent numbers, in a format suitable for re-submission as a URL.