Multivalued feature construction I have a task of representing a users feature matrix to be input for statistical machine learning classifiers, i have features like gender , age etc but I also have a multivalue feature called as "movies watched" which is essentially another table of movie names watched by that user with a numeric duration, the order of movies does not matter here. Also, movies watched can be from 20 movies to 300 movies. So what is the best way of representing this "movies watched" as a feature vector?
 A: 
I also have a multivalue feature called as "movies watched" which is essentially another table of movie names watched by that user with a numeric duration

Since duration is a continuous variable, a trivial solution would be to treat each movie as a separate continuous variable, where the duration varies from 0 to some upper bound, let's say 240 minutes (assuming movies in your database are not longer than 4 hours).
You can alternatively cluster your movies into meaningful semantic categories (e.g. horror, action, etc) and aggregate the time for each cluster. Then create a column for each cluster. This approach has an advantage that if your list of movies grows, you don't have to add additional column. If your matrix of users (rows) and movies (300 columns for 300 movies) is sparsely filled with continuous-valued data, consider applying Sparse PCA for dimensionality reduction. This does help reduce dimensionality but will be meaningful if semantic categories of the movies is well-covered by the 300 movies you have in your database.
