recommendation systems in mobile commerce presented by rachana chandrashekar(7487187)...

31
Recommendation Systems in Mobile Commerce Presented by Rachana Chandrashekar(7487187) [email protected]

Upload: nathan-curtis

Post on 16-Dec-2015

216 views

Category:

Documents


0 download

TRANSCRIPT

Recommendation Systems in Mobile Commerce

Presented byRachana Chandrashekar(7487187)

[email protected]

Your company slogan

1. INTRODUCTION

2. OVERVIEW

3. RECOMMENDER SYSTEM MODEL

OUTLINE

4. RECOMMENDATION ALGORITHMS

Your company slogan

6. CHALLENGES IN MOBILE COMMERCE

7. CONCLUSION

5. CHALLENGES OF RECOMMENDER SYSTEMS

OUTLINE

Your company slogan

INTRODUCTION

The Paradox of Choice

Overwhelming number of options to consider Lack of effective system support in making decisions Too many options can make your visitors too confused and undecided

Only 10% of products on an online retail store garner 75% of page views

Your company slogan

OVERVIEW

What are recommendation systems?

A recommendation system provides information or items that are likely to be

of interest to a user in an automated fashion.

Recommendation systems help match users with items

Your company slogan

WHY DO WE NEED RECOMMENDATION SYSTEMS?

Value for the Customer- Find things that are interesting- Narrow down the set of choices- Help explore the space of options- Reduce cognitive load on users

Value for the provider- Additional and unique personalized service for the customer- Increase trust and customer loyalty- Increase sales, click through rates etc.- Opportunities for promotion, persuasion- Obtain more knowledge about customers

Your company slogan

EVERYDAY EXAMPLES OF RECOMMENDATION SYSTEMS..

Your company slogan

MORE EXAMPLES..

Netflix predicts other “Movies You Love” based on past numeric

ratings (1-5 stars)

Recommendations drive 60% of Netflix’s DVD rentals

Your company slogan

MORE EXAMPLES..

Your company slogan

MORE EXAMPLES..

Your company slogan

RECOMMENDER SYSTEM MODEL

Candidate Generation

Rank

User Feedback

Filtering

Users Items

Automatically identify items of interest to users(Focus of talk)

Filters: near duplicates, already seen, dismissed

Recommendations based on temporal, geo-location and personalization

Track user feedback, likes, dislike, ratings

Your company slogan

RECOMMENDATION ALGORITHMS

Collaborative filtering (CF)

Hypothesis : Similar users tend to like similar items

Two forms of CF-Item-based collaborative filtering-User-based collaborative filtering

Data Collection Methods- Explicit feedback

Example: ratings, dismiss- Implicit feedback

Example: number of views, purchases

Your company slogan

DATA REPRESENTATION

Items : i1, i2, i3 …. in

User u1,u2,u3….un has provided ratings on items

Example of User/Movie Ratings Matrix:

Alice Bob Charlie Dave

Harry Potter … 3 5 2 3

American Pie 4 4 2 -

Twilight … 5 1 - -

Matrix - 1 1 5

Your company slogan

A NAÏVE RECOMMENDATION SYSTEM

1. Aggregate ratings for each item

2. Recommend item with maximum rating

score(i,u) = f(i) =

Does everybody like Harry Potter movies?

Historical information about users is important!

Your company slogan

Predict user’s rating for an item i based on his rating for other item Given a user u with I(u) preferred items

score(i,u) = sim(i,j)

Rating provided by user u for item j

Similarity between

items i and j

ITEM-BASED COLLABORATIVE FILTERING

Your company slogan

EXAMPLE : ITEM-BASED CF

Given user with ratings for items X and Y

Items N and S with similarities

score(u,N) = 1.0*0.8 + 0.3*0.3 = 0.89

score(u,S) = 0.2*0.8 + 0.3*0.8 = 0.4

Harry Potter (X) The Matrix (Y)

rating 0.8 0.3

Item Harry Potter (X) The Matrix (Y)

The Chronicles of Narnia (N)

1.0 0.3

Star Wars (S) 0.2 0.8

Your company slogan

COMPUTING SIMILARITY BETWEEN ITEMS

Cosine Similarity

- Items are represented as u-dimensional vectors over user space

- Similarity is cosine of the angle between two vectors

- Score ranges between 1 (perfect) and -1 (opposite)

Your company slogan

Example: 2 users

Items User 1 User 2

A 0.8 0.45

B 0.4 0.8

C 0.3 0.3

Your company slogan

JACCARD SIMILARITY MEASURE

Measures the similarity between finite sample sets

J(A,B) =

Defined as the size of intersection divided by the size of the union of the

sample sets

Sample sets of Items :

A ={Item1,Item3,Item6}

B ={Item1,Item2,Item6}

J(A,B) = = 0.5

Your company slogan

USER BASED COLLABORATIVE FILTERING

K – nearest neighbors ( KNN )- Group users into different clusters

Hard clustering Soft clustering

Users Clusters Items

Your company slogan

CONTENT-BASED RECOMMENDATION

Collaborative filtering does not require any information about the items

- However, it might be reasonable to exploit such information

-E.g. Recommend fantasy novels to people who liked fantasy novels in

the past

What do we need?

- Some information about the available items such as the genre

(content)

- Some sort of user profile describing what the user likes

Your company slogan

CONTENT REPRESENTATION AND ITEM SIMILARITIES

Your company slogan

HYBRID RECOMMENDER SYSTEMS

• Combination of collaborative filtering and content based

filtering

• Idea of crossing two or more implementations

• Hybrid features

- Social Features

Movies liked by user

- Content features

Dramas liked by user

- Hybrid features

User who like many movies that are dramas

Your company slogan

CHALLENGES AND INTERESTING PROBLEMS OF RECOMMENDER SYSTEMS

Data sparsity

-Users rarely purchase, rate or click

The more you see the less you know

- Increasing users or items increase the dimensions we need to learn

Cold-start problem

- No historical information for new users or items

Scalability

- Increase in the size of matrix

Your company slogan

CHALLENGES IN DESIGNING RECOMMENDER SYSTEMS FOR MOBILE USERS

• Size of the display, small screen devices

• Limited input and interaction capabilities

• Mobile users have shorter browsing sessions

• Lack of standardization of the browsing tools

• Cost of interaction

Exclusive characteristics :

• Location awareness

• Ubiquity

Your company slogan

CONCLUSION

Recommender systems are a huge success in E-commerce sites

Recommendation systems in mobile commerce have to overcome

obstacles

Mobile devices coupled with Recommender systems would be key

tools for business applications

Your company slogan

Question 1

In item based collaborative filtering, based on the user’s previous rating, recommend the most appropriate item to the user A.

Similarity with previously purchased items:

score (u,B) = 0.8*1.0 + 0.2*0.3 = 0.86

score(u,T) = 0.8*0 + 0.2*0.9 = 0.18

The item blueberry is recommended to the user as the score for blueberry is higher

User A Strawberries Oranges

Rating 0.8 0.2

Item Strawberries Oranges

Blueberry (B) 1.0 0.3

Tangerine (T) 0 0.9

Your company slogan

Question 2

In user based collaborative filtering, using Jaccard Similarity find the

similarity between two users based on the books that they prefer.

Determine whether users are similar or not.

Users = { User A, User B }

A = { The Hobbit, Harry Potter and the Deathly Hallows , Angels and

Demons }

B = { Angels and Demons, Digital Fortress, The Lost Symbol }

J(A,B) =

= = 0.2 Since the score is nearing zero, users are dissimilar

Your company slogan

Question 3

Using hybrid recommendation( both collaborative and content based filtering) predict the top 3 movie recommendations for user Karen. In the below problem, given is a set of users with a set of their preferred movies belonging to different genres. New User Karen likes Amelie. Based on this data, predict the next 3 recommendations for Karen. Set of Users = {Brian, Ellen, Fred, Dean, Jason}Set of Movies = {Amelie, Star Wars, Hiver, Whispers, Batman, Rambo}Genre = {Action=(Batman, Rambo), Foreign=(Amelie, Hiver, Whispers), Classic=(Star Wars)}

Users Movies

Brian Amelie Star Wars

Ellen Amelie Star Wars Hiver

Fred Star Wars Batman

Dean Star Wars Batman Rambo

Jason Hiver Whispers

Karen ? ? ?

1. Star Wars2. Hiver3. Whispers

Your company slogan

New User Karen likes Amelie.

Based on this data, look for users who like the same movie. Brian and Allen are the two other users who like Amelie. Both of them also

like Star Wars. So Star Wars would be the first movie to be recommended to Karen based on user-item similarity (Collaborative filtering)

User Ellen who likes Amelie and Star Wars also likes Hiver. So Hiver would be the next movie to be recommended to Karen.

After recommending Hiver, now we look at users who like Hiver ( Hiver belongs to genre foreign )

User Jason likes Hiver and Whispers. Hiver and whispers belong to genre – foreign. Now these movies can be matched to user Karen’s original liked movie Amelie ( genre – foreign). Based on content based filtering ( genre)

the next movie recommended to Karen is Whispers.

Thus the top three movie recommendations to user Karen are Star Wars, Hiver and Whispers.

Your company slogan

REFERENCES

• Chengzhi Liu, Caihong Sun and Meiqi Fang, The design of an open hybrid recommendation system for mobile commerce, Communication Technology, 2008. ICCT 2008. 11th IEEE International Conference on E-ISBN: 978-1-4244-2251-7

• Azene Zenebe, Ant Ozok and Anthony F. Norcio, Personalized Recommender Systems in e-commerce and m-commerce:A comparitive Study,11th International Conference on Human-Computer Interaction

• Badrul Sarwar, George Karypis, Joseph Konstan and John Riedl, Analysis of recommendation algorithms for e-commerce, EC '00 Proceedings of the 2nd ACM conference on Electronic commerce, ISBN:1-58113-272-7

• Amund Tveit, Peer to peer based Recommendation for mobile-commerce, ACM Mobile Commerce Workshop,2001