Amazon : Determine which books are going to sell or not using k-means clustering and Amazon AWS

Changes

1. Initial Version

Installation

1. install Amazon's AWS, get a seller id, key id, and secret key id
2. under cygwin set up ~/.amazonrc as directed by ruby-aws
3. run gem install ruby-aaws
4. modify the line for "seller_id = " in amazon_gen.rb with YOUR seller id

Notes

If you sell Amazon books, these scripts will help you predict which books will sell and which won't.

To use

1. Do these steps ONCE:
   • run ruby amazon_gen.rb to generate "books.rb"
   • rename it to "mybooks.rb"
2. run ruby amazon_kmeans.rb > kmeans.out to generate a list of books and clusters
3. as you sell books, edit mybooks.rb, changing "N" to "Y" for those books you sold
4. re-run ruby amazon_kmeans.rb > kmeans.out

Notes

• amazon_kmeans.rb takes a long time to run. On my system (2.8Ghz Quad) with under 300 books in "books.rb", it takes more than seven minutes to run.
• You can re-generate the books.rb file, but you will have to merge the ones that are sold into that file. The AWS query only generates entries for those that are unsold.
• The clusters are stable. Once a book is allocated to a cluster it tends to stay within that cluster. In my system, after more than four months of use, no books have moved from one cluster to another.
• The clusters seem to be heavily related to the Amazon ranking and have little to do with price or number of books offered.

Interpreting the results

In file "kmeans.out", look for lines like:

    ---- Number of titles in cluster 9: 77  #sold=13

This line, on the other hand,

    ---- Number of titles in cluster 8: 32  #sold=0

The other lines will be a dump of your books:

  9 : [4.0, 46, 0.2, 715938, "N"] N - About Face: The Essentials of User Interface Design by Cooper, Ala
  9 : [4.0, 25, 3.4, 714208, "Y"] Y - Boris and Bella [Hardcover] by Crimi, Carolyn; Grimly, Gris

The book "About Face" has not sold. It has an Amazon rank of 715,938. I am selling it for $4.00 and the lowest price for other books is $0.20. There are 46 other sellers for this book.

The book "Boris and Bella" has sold. It has an Amazon rank of 714,208. I was selling it for $4.00 and the lowest price for other books (at the time!!) was $3.40. There were 25 other sellers for this book.