Netflix, While I Am Waiting, ...

While I am waiting for the full set of qualifying.txt, I spent some time in analysing the raw data. 90-20 rule applies to the training_set as well as the qualifying data. That's why I started my buddy building from the less rated movies.

Although my run has completed 50% of movies, it only finished 4% of the total predictions.

I also wrote an AWK program to summarise the ratings (1-5) based on all the customers. Do you know that 1269 customers rated 1 time and 16419 customers rated less than 10 times, considered the average ratings per customer is 209. As for the top rated customer #305344, he/she rated 17653 movies over a 6 years period (320 weeks). That means he/she rated 55 movies per week, possible ? Obviously he/she cannot be possible to watch that many movies. BTW, we are talking about 'rated', not 'watched'.