mahout - recommendation system data collection methodology -

I am creating a recommendation system in my application and I'm probably going to use Apache Mahavat, I Large dataset, it will be collected over a period of time ... When creating any type of log file collected in a DB, it will be the least expensive to earn and I will need it if I export it. / P>

Recommendation of Mahavit The code can be read directly from a database or file - this normal log files will not be read if the data is formatted correctly; They should be translated into simple CSV or TSV but it can read about any table that contains user / item / preferences.

If you are already putting your data in the database table, then I would say that it should be left there and do not duplicate it or export it worthless. If possible, you should remember the fragments in memory. Will have to suck

If you are not already collecting this data, and want to choose a simple and efficient representation, then I suggest that you remove the user / items / priority information and give them simple csv files Stored in, compress with gzip. It can be used easily with the festival and the whole log will be more simple and more compact than a file or database.