tangent : a novel, “surprise-me”, recommendation algorithm

Intelligent Database Systems Lab

國立雲林科技大學National Yunlin University of Science and Technology

1

TANGENT: A Novel, “Surprise-me”, Recommendation Algorithm

Kensuke Onuma , Hanghang Tong , Christos Faloutsos2009.SIGKDD

Presented by Chien-Hao Kung2011/8/10

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

2

Outlines· Motivation· Objectives· Methodology· Experiments· Conclusions· Comments

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

3

Motivation

· Most of the recommendation algorithms focus on the precision in the proximity to user preferences. However, this strategy tends to suggest items only on the center of user preferences and thus narrows down the users’ horizons.

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

4

Objectives

· To propose a method which are well connected to older choices, and at the same time well connected to unrelated choices.

· The method is carefully designed to be (a) parameter-free (b) effective and (c) fast.

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Methodology

Define TANGENT problem as follows: Given: an edge-weighted undirected graph G with

adjacency matrix A, the set of query nodes Q=()1i<k. Find: a node that (1) is close enough to Q, and (2) has

high potential to reach other nodes.

5

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Methodology

Framework of TANGENT Algorithm Step1: Calculate relevance score (RS) of each node: Step2: Calculate Bridging Score (BRS) of each node: Step3:Compute the TANGENT score () by somehow

merging two criteria above.

6

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Methodology

Relevance Score (RS) It’s proposed to use random walk with restart.

7

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Methodology

Bridging Score (BRS)

8

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Methodology

TANGENT Score Method 1: To use linear combination.

= Method 2: Skyline queries. Proposed Combination Method.

=

9

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Methodology

Scalability Computing Relevance Score

Using random walk with restart Computing Bridging Score

The R can be re-used in computing bridging score It doesn’t need to compute bridging scores of user

nodes for recommendation. Merging

It needs just a multiplication for each of the n-q=O(n) candidate nodes.

10

=

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

Synthetic Data Sets

Experiments

11

=1.95

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

MovieLens Data Set(Slapstick Movie Fan’ case )

Experiments

12

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

MovieLens Data Set(Horror Movie Fan’ case )

Experiments

13

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

MoveLens data setExperiments

14

=0.08

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

CIKM data setExperiments

15

=0.22

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.Experiments

DBLP Data Set

16

𝑆𝑢𝑟𝑝𝑟𝑖𝑠𝑖𝑛𝑔 𝐺𝑎𝑖𝑛 :𝑛𝑡𝑎𝑛−𝑛𝑟𝑒𝑙

𝑛𝑡𝑎𝑛

0.780.70

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

17

Conclusions· It’s proposed TANGENT algorithm to find items that are close

to the user preferences, while they also have high connectivity to other groups.

· Careful design decisions, so that the resulting method is (a) parameter-free (b) effective and (c) fast.

Intelligent Database Systems Lab

N.Y.U.S.T.

I. M.

18

Comments

· Advantages─ there are many pictures in this paper, so it can be

read intuitively

· application─ Information Storage and Retrieval