algoGod Frequently Asked Questions

Q. What is the database md5sum?
A. The database size and md5sum are following:
ls -lt
rw-r--r-- 1 root root 144996092 Oct 15 22:57 AlgoGod.sql.gz
# md5sum AlgoGod.sql.gz
344647429a256e7d471f52d00a634af2 AlgoGod.sql.gz

Q. In the problem statement, you have mentioned that “The columns marked in blue are ‘features’ that could be used for machine learning. I request you to confirm that this is just a suggestion, and that we need not limit ourselves to using only those particular columns for machine learning.
A.The list of features is just a suggestion. You can use any number of features.

Q. There are only 8 distinct user ip addresses. There must be many more useres than that. How do you develop an algorithm to put ad for individual user?
A. You can use user_guid also as a feature.

Q. What exactly the user guid is?
A. The user_guid uniquely identifies a user.

Q. There are many entries in the table where click count is greater than 1. What exactly the click count means.
A. Please ignore clicks more than 1. Anything more than 1 is equal to one.

Q. Is the predictions just based on the colors going to yield good results.
A. Yes, color optimization has shown up to 50% improvement in the CTR. We think we can do better.

Q. Does "Yin-Yang" algorithm makes use of only these attributes?
A. Yes, Yin-Yang only uses these attributes.

Q. When we do develop the algorithm or try any composite methods for predicting the click through rate, is there any measure by which the accuracy could be determined?
A. At this point we don’t have a publicly available mechnism to measure the accuracy of your algorithm. You can submit more than 1 algorithm, if you believe that an alternative algorithm will get better results.



More questions? Send email to algogod@komli.com .