Slope One Predictors for Online Rating-Based Collaborative Filte...
5星 · 超过95%的资源 需积分: 9 77 浏览量
2013-09-06
10:12:19
上传
评论
收藏 66KB PDF 举报 
Slope One Predictors for Online Rating-Based Collaborative Filtering
Daniel Lemire
∗
Anna Maclachlan
†
February 7, 2005
Abstract
Rating-based collaborative filtering is the process of predict-
ing how a user would rate a given item from other user
ratings. We propose three related slope one schemes with
predictors of the form f(x) = x + b, which precompute the
average difference between the ratings of one item and an-
other for users who rated both. Slope one algorithms are
easy to implement, efficient to query, reasonably accurate,
and they support both online queries and dynamic updates,
which makes them good candidates for real-world systems.
The basic SLOPE ONE scheme is suggested as a new ref-
erence scheme for collaborative filtering. By factoring in
items that a user liked separately from items that a user dis-
liked, we achieve results competitive with slower memory-
based schemes over the standard benchmark EachMovie and
Movielens data sets while better fulfilling the desiderata of
CF applications.
Keywords: Collaborative Filtering, Recommender, e-
Commerce, Data Mining, Knowledge Discovery
1 Introduction
An online rating-based Collaborative Filtering CF query
consists of an array of (item, rating) pairs from a single user.
The response to that query is an array of predicted (item,
rating) pairs for those items the user has not yet rated. We
aim to provide robust CF schemes that are:
1. easy to implement and maintain: all aggregated data
should be easily interpreted by the average engineer and
algorithms should be easy to implement and test;
2. updateable on the fly: the addition of a new rating
should change all predictions instantaneously;
3. efficient at query time: queries should be fast, possibly
at the expense of storage;
4. expect little from first visitors: a user with few ratings
should receive valid recommendations;
∗
Université du Québec
†
Idilia Inc.
In SIAM Data Mining (SDM’05), Newport Beach, California, April
21-23, 2005.
5. accurate within reason: the schemes should be compet-
itive with the most accurate schemes, but a minor gain
in accuracy is not always worth a major sacrifice in sim-
plicity or scalability.
Our goal in this paper is not to compare the accuracy
of a wide range of CF algorithms but rather to demonstrate
that the Slope One schemes simultaneously fulfill all five
goals. In spite of the fact that our schemes are simple,
updateable, computationally efficient, and scalable, they are
comparable in accuracy to schemes that forego some of the
other advantages.
Our Slope One algorithms work on the intuitive prin-
ciple of a “popularity differential” between items for users.
In a pairwise fashion, we determine how much better one
item is liked than another. One way to measure this differen-
tial is simply to subtract the average rating of the two items.
In turn, this difference can be used to predict another user’s
rating of one of those items, given their rating of the other.
Consider two users A and B, two items I and J and Fig. 1.
User A gave item I a rating of 1, whereas user B gave it a
rating of 2, while user A gave item J a rating of 1.5. We ob-
serve that item J is rated more than item I by 1.5 − 1 = 0.5
points, thus we could predict that user B will give item J a
rating of 2+0.5 = 2.5. We call user B the predictee user and
item J the predictee item. Many such differentials exist in a
training set for each unknown rating and we take an average
of these differentials. The family of slope one schemes pre-
sented here arise from the three ways we select the relevant
differentials to arrive at a single prediction.
The main contribution of this paper is to present slope
one CF predictors and demonstrate that they are competitive
with memory-based schemes having almost identical accu-
racy, while being more amenable to the CF task.
2 Related Work
2.1 Memory-Based and Model-Based Schemes
Memory-based collaborative filtering uses a similarity
measure between pairs of users to build a prediction,
typically through a weighted average [2, 12, 13, 18]. The
chosen similarity measure determines the accuracy of the
prediction and numerous alternatives have been studied [8].
Some potential drawbacks of memory-based CF include
资源评论
wolf616002017-08-08学习协同推荐算法的童鞋可以看，很不错
wangyuquan2015-03-24论文介绍了三种方法，感觉很不错
