1 RegionKNN: A Scalable Hybrid Collaborative Filtering Algorithm for Personalized Web Service Recommendation Xi Chen, Xudong Liu, Zicheng Huang, and Hailong.

1

RegionKNN: A Scalable Hybrid Collaborative Filtering Algorithm for Personalized Web

Service Recommendation

Xi Chen, Xudong Liu, Zicheng Huang, and Hailong Sun

School of Computer Science and EngineeringBeihang University

Beijing, China

2

Outline

• Introduction

• Motivation

• RegionKNN Algorithm

• Experiments

• Conclusion and Future Work

3

1. Introduction

4

Introduction

• Current situation– More than 25,000 public available services (seekda.com)– About 200,000 related documents

• Goal of service recommendation– Optimal QoS– User preference

• Current method: Collaborative Filtering (CF) – predict and recommend the potential favorite items for a

particular user by using rating data collected from similar users.

• If Alice and Bob both like X and Alice likes Y then Bob is more likely to like Y

• Problems– Characteristics of QoS are neglected– Online performance need to be improved

5

2. Motivation

6

A Motivating Scenario

Some QoS properties (e.g. availability, response time) highly correlate to users’ physical locations.

EmailFiltering

WS

EmailFiltering

WS

7

3. RegionKNN Algorithm

8

What’s RegionKNN• Hybrid CF Algorithm

– recommend web services with optimal QoS to the active user with consideration of the region factor

• Two phases of RegionKNN– Region model building (offline)

• Region-sensitive services identification

• Region aggregation

– Service recommendation (online) (modified KNN) • Neighbor selection

• QoS Prediction

I take response time/round trip time (RTT) as an example to describe our algorithm

9

3.1 Region model

10

Region Model• Region

– a group of users who are closely located with each other and have similar RTT values

u5

u19 u2u22,u8

u1, u3

Service A Service A

Service B Service B

Service X Service X

11

Input Dataset

• User-Service RTT Matrix: m services, n users

• The set of non-zero RTTs of service s {R1(s), R2(s),…, Rk(s)} collected from all users is a sample from population R.

s1 s2 … sm

u1 0 245 … 20078

u2 2023 342 … 539

… … … … …

un 0 3040 … 498

RTT is much longer than

others

12

Region-sensitive Services Identification

• To estimate the mean μ and the standard deviation σ of R, we use:

))((ˆ sRmedian ii

))((4862.1ˆ sRMAD ii

Median: the numeric value separating the higher half of a sample from the lower half. e.g. {120, 128, 200, 250, 258, 2000, 3500} median = 250

MAD: the Median of the Absolute Deviations from the sample's median. e.g. {120, 128, 200, 250, 258, 2000, 3500} {8, 50, 122, 130, 1750, 2250} MAD = 130

13

• Region-Sensitive Service– Let R = {R1(s), R2(s),…, Rk(s)} be the set of RTTs of service s

provided by users from all regions. Service s is a sensitive service to region M iff

))()ˆ3ˆ)((()( MjregionsRRsR jj

{120, 128, 200, 250, 258, 2000, 3500}

u1 u3 u5 u19 u2 u22 u8

u5

u19 u2

u22

u8u1, u3

Service A

Service A

Region-sensitive services Identification

14

Definition

• Region Sensitivity

• Sensitive Region– Region M is a sensitive region iff regSen >λ.

• Region center – the median vector of all the RTT vectors provided by

users in a region

||

||)(

services

ervicessensitivesMregSen

15

Region Aggregation

• Why?– Users only provide limited number of QoS values, the

sparse dataset always leads to poor recommendation.

• How?– It treats users with similar IP addresses as a region at

the outset – In each iteration, the two most similar and non-sensitive

regions are selected and aggregated, if their similarity exceeds threshold μ.

– It executes at most N-1 steps (N is the number of regions at the outset), in case that all regions are non-sensitive, extremely correlates to each other and finally aggregates into one region.

16

Region Similarity• The similarity between region M and N is measured by

the similarity of the two centers. • Similarity by Pearson Correlation Coefficient (PCC)

)()(

2

)()(

2

)()(

))(())((

))(())((

),(

mSnSsnn

mSnSsmm

nnmSnSs

mm

RsRRsR

RsRRsR

nmSim

s1 s2 s3 s4 s5

cm 1 2 5 0 0

cn 0 0 5 1 3

By PCC, the similarity is of the two regions is 1

17

Region Similarity

• PCC often overestimates the similarity when the two regions have few co-invoked services. To adjust it, we use:

),(|)()(|

|)()(|),( nmSim

nSmS

nSmSnmmSi

s1 s2 s3 s4 s5

cm 1 2 5 0 0

cn 0 0 5 1 3

By adjustment, the similarity of the two regions is 0.2

18

3.2 Service Recommendation

19

Neighbor Selection

• Neighbors: users with similar QoS experiences• Advantages of region-based neighbor selection

– Do not need to search the entire dataset, thousands of users are clustered into a certain number of regions

– The feature of the group of users in a region is represented by the region center

20

QoS Prediction

• To calculate the RTT prediction for the active user u and service si

• Get the active user’s IP address and find the region the user belongs to. If no appropriate region is found, the active user will be treated as a member of a new region.

• Identify whether service si is sensitive to the specific region. If it is region-sensitive, then the prediction is generated from the region center:

)(ˆ iu sR

)()(ˆ icenteriu sRsR

21

QoS Prediction (cont.)

• Otherwise, use adjusted PCC to compute the similarity between the active user and each region center that has evaluated service si, and find up to k most similar centers {c1, c2,…, ck}.

• If the active user’s region center has the RTT value of si, i.e. , the prediction is computed using the equation:

0)( icenter sR

k

j j

k

j jcic

icenteriucumSi

cumSiRsRsRsR jj

1

1

),(

),())(()()(ˆ

22

QoS Prediction (cont.)• Otherwise,

• Previous CF-based web service recommendation

algorithms use the following equation, to predict the missing QoS value.

• This equation is based on the assumption that each user’s rating range is subjective and comparatively fixed, while it

is not applicable in our context.

k

j j

k

j jic

iucumSi

cumSisRsR j

1

1

),(

),()()(ˆ

k

j j

k

j jcic

uiucumSi

cumSiRsRRsR jj

1

1

),(

),())(()(ˆ

23

Time complexity

• Model building (offline)– The time complexity of region aggregation algorithm

is O(N2logN), and N is the number of regions at the outset.

• QoS prediction (online)– Let l be the number of regions, m the number of web

services, and n the number of users. In the online part, O(l) similarity weight calculations are needed, each of which takes O(m) time. Therefore, the online time-complexity is O(lm)≈O(m). Previous user-based CF algorithm has O(mn) online time complexity.

24

4. Experiments

25

Experiments

• Dataset– a subset of WSRec with 300,000 RTT records– 3000 users – 100 services

• Evaluation Metric

– Ru(s) denotes the actual RTT of web service s given by user u

– denotes the predicted one– L denotes the number of tested services

L

sRsR

MAE suuu

,

|)(ˆ)(|

)(ˆ sRu

Dataset: http://www.wsdream.net

26

MAE Performance

27

Impact of λ and μ

28

Impact of neighborhood size K

29

Impact of Data Sparsity

30

5.Conclustion and Future Work

31

Conclusion and Future Work

• Conclusion– a new region model for clustering users and

identifying region-sensitive web services– a hybrid model-based and memory-based CF

algorithm for web service recommendation, which significantly improves the recommendation accuracy

– We demonstrate RegionKNN’s scalability advantage over traditional CF algorithms via time-complexity analysis

• Future Work– Investigation of more QoS properties and their

variation with time – Internal relations between QoS properties

32