Recommendation systems

IMDAMohammad Hamidi Esfahani
Recommendation Systems
‫اصفهانی‬ ‫حمیدی‬ ‫محمد‬
‫مردادماه‬1394

Introduction
Recommender systems are widely used on the Web
for recommending products and services to users.
Most e-commerce sites have such systems.
These systems serve two important functions.
They help users deal with the information overload by giving
them recommendations of products, etc.
They help businesses make more profits, i.e., selling more
products.

Ecommerce Concerns
How to collect more users ???
How to settled the users in our website ???
Which product to sell or resell ???
Which users are important ????

Recommendation Systems
When ???
Which ???
How ???
FeedBack
Improve Recommendation

Facts
Rating
Additional attributes
Likelihood
Item Similarity
User Similarity
Information Gathering
Biggest Problem: Cold Start

Recommendation Systems(Conti..)
Recommendation Method
Collaborative
Content-Based
Knowledge-Based
Hybrid

Collaborative Method
if user A and user B have a
purchase history that overlaps
strongly and user A has
recently bought an item that B
has not yet been, the basic
rationale is to propose this
item also to B

Maintain a database of many users’ ratings of a variety
of items.
For a given user, find other similar users whose ratings
strongly correlate with the current user.
Recommend items rated highly by these similar users,
but not rated by the current user.
Almost all existing commercial recommenders use this
approach (e.g. Amazon).

A 9
B 3
C
: :
Z 5
A
B
C 9
: :
Z 10
A 5
B 3
C
: :
Z 7
A
B
C 8
: :
Z
A 6
B 4
C
: :
Z
A 10
B 4
C 8
. .
Z 1
User
Database
Active
User
Correlation Match
A 9
B 3
C
. .
Z 5
A 9
B 3
C
: :
Z 5
A 10
B 4
C 8
. .
Z 1
Extract
Recommendations
C

Weight all users with respect to similarity with the
active user.
Select a subset of the users (neighbors) to use as
predictors.
Normalize ratings and compute a prediction from a
weighted combination of the selected neighbors’
ratings.
Present items with highest predicted ratings as
recommendations.

Typical Questions
How to find users with similar tastes to the user for whom we
need a recommendation?
How to measure similarity?
What should be done with new users, for whom a buying
history is not yet available?
How to deal with new items that nobody has bought yet?
What other techniques besides looking for similar users can
use for making a prediction about whether a certain user will
like an item?

Content Based Method
if item A and item B have similar
attributes, the basic rationale is
to Recommend item B to whom
that buy item A

Content Based Method (Conti..)
The key idea of content-based filtering approach is
based on similarity of item’s features.
The best idea in content-based recommendation
system, is to recommend similar items to user that
buy the first item previously.

Content Based Method
Recommendations are based on information on the
content of items rather than on other users’ opinions.
Uses a machine learning algorithm to induce a profile of
the users preferences from examples based on a featural
description of content.
Some previous applications:
Newsweeder (Lang, 1995)
Syskill and Webert (Pazzani et al., 1996)

Advantages of Content-Based Approach
No need for data on other users.
No cold-start or sparsity problems.
Able to recommend to users with unique tastes.
Able to recommend new and unpopular items
 No first-rater problem.
Can provide explanations of recommended items by
listing content-features that caused an item to be
recommended.
15

Disadvantages of Content-Based Method
Requires content that can be encoded as
meaningful features.
Users’ tastes must be represented as a learnable
function of these content features.
Unable to exploit quality judgments of other users.
Unless these are somehow included in the content
features.
16

Typical Questions
How can systems automatically acquire and
continuously improve user profiles?
How to determine which items match, or are at least
similar to or compatible with, a user’s interests?
What techniques can be used to automatically
extract or learn the item descriptions to reduce
manual annotation?

Knowledge
Based
(Rule Based)
Typical customers buy a new camera only once
every few years, so the recommender system
cannot construct a user profile or propose
cameras that others liked, which would result in
proposing only top-selling items.

Knowledge Based Method
The recommender system typically makes use of
additional, often manually provided, information
about both the current user and the available items
User Interaction
Direct
Indirect

Typical Questions
 What kinds of domain knowledge can be represented in a knowledge
base?
 What mechanisms can be used to select and rank the items based on the
user’s characteristics?
 How to acquire the user profile in domains in which no purchase history is
available, and how to take the customer’s explicit preferences into
account?
 Which interaction patterns can be used in interactive recommender
systems?
 Finally, in which dimensions can personalize the dialog to maximize the
precision of the preference elicitation process?

Hybrid Approaches
One obvious solution is to
combine different techniques
to generate better or more
precise recommendations

Typical Questions
Which techniques can be combined, and what are
the prerequisites for a given combination?
Should proposals be calculated for two or more
systems sequentially, or do other hybridization
designs exist?
How should the results of different techniques be
weighted and can they be determined dynamically?

Sample Hybrid Method
Proposed hybrid recommendation system method
Knowledge-
basedpart
Collaborative
part
Content-basedpart
User
interaction-
basedpart
Item’s
features
User’s
properties
C-Means Clustering Method
Extract Weighted Map
Users
History
Extract Rules
Prioritize Rules
Users Interaction
Like/Dislike

Proposed Method (Conti..)
First off all, Extract the clusters on items and users
based an their profile features and the belonging
factor for each item or user
Second, Extract the relations and relation weight
between clusters

Proteins
Classifying VS Clustering
 Objects characterized by one or more features
 Classification
 Have labels for some points
 Want a “rule” that will accurately assign labels to new
points
 Supervised learning
 Clustering
 No labels
 Group points into clusters based on how “near” they
are to one another
 Identify structure in data
 Unsupervised learning
Genes

Similarity
KNN (K-Nearest Neighbor)
K-Means
C-Means

K-Nearest Neighbor
Features
All instances correspond to points in an n-dimensional
Euclidean space
Classification is delayed till a new instance arrives
Classification done by comparing feature vectors of the
different points
Target function may be discrete or real-valued

1-Nearest Neighbor

3-Nearest Neighbor

K-Nearest Neighbor
An arbitrary instance is represented by (a1(x), a2(x), a3(x),..,
an(x))
ai(x) denotes features
Euclidean distance between two instances
d(xi, xj)=sqrt (sum for r=1 to n (ar(xi) - ar(xj))2)
Continuous valued target function
 mean value of the k nearest training examples

The K-Means Clustering Method
 Given k, the k-means algorithm is implemented in four
steps:
1. Partition objects into k nonempty subsets
2. Compute seed points as the centroids of the clusters of the current partition
(the centroid is the center, i.e., mean point, of the cluster)
3. Assign each object to the cluster with the nearest seed point
4. Go back to Step 2, stop when no more new assignment

K-means Clustering

Fuzzy C-means Clustering
Fuzzy c-means (FCM) is a method of clustering
which allows one piece of data to belong to two or
more clusters.
This method (developed by Dunn in 1973 and
improved by Bezdek in 1981) is frequently used in
pattern recognition.

Fuzzy C-means Clustering

item 𝐼 𝑎 belongs to cluster 𝐶𝑖 by 𝐵(𝐶𝑖𝑎) degree
Item 𝐼 𝑏 belong to cluster 𝐶𝑖 by 𝐵(𝐶𝑖𝑏) degree
User 𝑈 𝑎 belongs to cluster 𝐶 𝑢 by 𝐵(𝐶 𝑢𝑎) degree
User 𝑈 𝑏 belongs to cluster 𝐶 𝑢 by 𝐵(𝐶 𝑢𝑏) degree
𝑈 𝑎 buy 𝐼 𝑎 and user 𝑈 𝑏 don’t buy any item and item
𝐼 𝑏 never buy

 cluster 𝐶𝑖 on items has a relation to cluster 𝐶 𝑢 on users by 𝑀(𝐶𝑖𝑢)
 𝑀 𝐶𝑖𝑢 =
𝑥
𝑖𝑡𝑒𝑚𝑠
𝑦
𝑢𝑠𝑒𝑟𝑠 𝐵 𝐶 𝑢𝑥 ∗ 𝐵(𝐶 𝑖𝑦)
𝑚∗𝑛
 m => the count of users
 n => the count of items.
 Then, a rule generated for recommending items in cluster 𝐶𝑖 on items to
users in cluster 𝐶 𝑢 on users. So, user 𝑈 𝑏 have a relation with item 𝐼 𝑏
 𝑅𝑒𝑐 𝑈 𝑏, 𝐼 𝑏 = 𝑀 𝐶𝑖𝑢 ∗ 𝐵 𝐶𝑖𝑏 ∗ 𝐵 𝐶 𝑢𝑏

SahamKaav.iR
proposed
method
with c-means
proposed
method
with k-means
random
recommendation
Number of
clusters of users
14 10 -
Number of
clusters of Stocks
56 48 -
LF (Like Factor ) 78% 53% 32%

Q & AMohammad Hamidi Esfahani
m.hamidi.es@gmail.com
Twitter: @haj_mamed

Recommendation systems

Recommended

Recommended

More Related Content

Similar to Recommendation systems

Similar to Recommendation systems (20)

Recently uploaded

Recently uploaded (20)

Recommendation systems