sied project when reading book Collabrorative Filting
In this project, we build a recommendation system for different users. We use user-based collabrative filtering to rank the related goods for selected custermer.
- First we calculate the similarities of all the other users.
- Then we obtain the weighted sum of the ratings of related goods over all people using weighted_sum = \sum rating_of_goodd * simlarity_of_that person
- Finally we get the scores for each good and recommend them by reverse order using score = weighted_sum/\sum_of_similarities
Further, we can also use item based filtering (switching item with person) to do recommendations - faster in digital marketing and used by Amazon