Take the 2-minute tour ×
Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

I am trying to mine product-usage sequences for multiple users of online gaming site. I have found the R package arulesSequences but am not sure how to fit it to my problem. The data format would be tuples similar to those used in arulesSequences, but instead of just mentioning products A and B in transactions, I would like to mention the quantities in which those products (games of different types) were bought (played, in my case).

Example:

sequence1 (for uid=1)

Date   UID Game1     Game2      Game3     Game4
Jan1    1   125times 0times     0times     0times
Jan2    1   0times   1time      0times     0times

Each user would have such a sequence. However, I see that arulesSequences function cspades would only allow to operate with bollean types, e.g. whether each game was actually played on that date:

Example:

sequence1 (for uid=1)

Date   UID 
Jan1    1   Game1 Game2
Jan2    1   Game1
Jan3    1   Game2 Game3

My goal is to determine rules like "if a user playes Game3 on that date, it causes them to play much more of Game2 one week later".

share|improve this question
add comment

2 Answers

I don't think that arulesSequences is really appropriate in your case, because it essentially focuses on 1-dimensional sequences.

I would suggest a model relating the usage of a game i in time t as a function of the usage of games 1-4 in time t-1, t-2, etc. A regression should be appropriate.

share|improve this answer
    
thank you for the answer. I have been so fixated on using association rules (or sequence clustering) with this problem that I have not considered to try other techniques. I have searched through all the packages and haven't found anything. Will try regression. –  zima Oct 15 '12 at 10:53
add comment

Could this be best modeled as a time series analysis? The counts for each game form a time series and your goal is to see if you can predict a series either by itself or by using all other series as a predictor.

But each day's count doesn't seem to be independent of the previous day's count which I why I thought a time series was most suitable.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.