We know where you should work next summer: job recommendations

We know where you

should work next

summerJob Recommendations

September 2015 @RecSys

Fabian Abel, http://xing.com

We know where you

should work next

summerJob Recommendations

September 2015 @RecSys

Fabian Abel, http://xing.comAt XINGin Hamburg, Barcelona, Vienna

or somewhere remote…

ChallengeIdentifying job postings that match the demands of the user and employer

3

Job

Recommender

0.92 0.8 0.76

…

User

Job postingsEmployer

Job recommendations

Job recommendations

6

explanations feedback

Job postings on XING

7

Title

Company

Employment type

and career level

Full-text

description

Key properties of a job posting

Key sources for understanding user demandsExploiting patterns that are found in the data(graph)

8

Social Network

explicit and

implicit

connections

Profile

Fabian Abel

Data Scientist

Haves:

Interests:

web science

big data, hadoop skills & co.

Interactions

data

web

social media

clicks, shares,

ratings

big data

kununu

Interactions of

similar users

similar usershadoop

scala

Social Network

explicit and

implicit

connections

Profile

Fabian Abel

Data Scientist

Haves:

Interests:

web science

big data, hadoop skills & co.

Interactions

data

web

social media

clicks, shares,

ratings

big data

kununu

Interactions of

similar users

similar usershadoop

scala

Relevance EstimationFinal relevance score of an item is obtained by combining the

scores coming from the “sub-recommenders” (= features)

9

Content-

based

features

Collaborative

features

Social

features

Usage

behavior

features

Relevance

Estimation(regression model)

Logistic Regression

P(relevant | x) = 1

1 + e -(b0 + bi xi)i

n

feature vector impact of feature xi

Relevance Estimation + Additional FiltersFiltering (rules) may dampen the relevance scores or filter out items

10

Content-

based

features

Collaborative

features

Social

features

Usage

behavior

features

Relevance

Estimation(regression model)

Location-

based

filtering

Content-

based

diversification

Monetary-

based

diversification

Career Level

filtering

Filtering &

Diversification

0.92 0.8 0.76

…

past

past

Profile describes a

user‘s past/current

position(s), not her

future career step(s)

Career path patterns: locationsDistance between user and location of bookmarked job postings on XING

11

0-50 km

35%

51-200 km

22%

>200 km43%

Career path patterns: career levelsClimbing up the ladder (based on 15M XING CVs)

12

junior

junior

senior

manger

senior manger

today

Next

ste

p

53%

senior

72%

manger

54%

senior manger

52%

Career path patterns: job rolesMost users switch at least once from one job role to another

13

Postdoc

Manager

Lecturer

Postdoc

Professor

6%

5%

3%

2%

Career path transitionsUnderstanding transitions in the career path graph

14

Web Developer

J2EE Developer

Data Scientist

Machine Learning Expert

MSc Computer Science

CV:

MSc Computer

Science

Web Developer

J2EE

Developer

Data Scientist

Machine

Learning

Expert

Data Scientist

Machine Learning Expert

PhD Data Mining

CV:

J2EE Developer

Data Scientist

Machine Learning Expert

MSc Computer Science

CV:

PhD Data

Mining

Career path graphWeighted directed graph with different types of nodes (job roles, education)

15

Association rule mining for constructing

the career path graph:

• Association rules (= edges):

Job role A Job role B

Education X Job Role Y

...

• Minimum support (e.g. at least k

transitions with A and B have to occur in

the data)

• Minimum confidence (= probability(B | A)

= weights of edges)

MSc Computer

Science

Web Developer

J2EE

Developer

Data Scientist

Machine

Learning

Expert

PhD Data

Mining

Similarly, graphs are constructed for:Jobrole X Industry Y

Career Level X Career Level Y...

Thresholds for min-support and min-confidence need to be learned (per

“discipline”)

Inferring Features from Career path graph(s)Probabilities that the job role is appropriate for the user

16

User

Machine

Learning

Expert

PhD Data

Mining

Job posting

Data

Scientist

P( | , )Data

Scientist

Machine

Learning

Expert= 0.79F2:

PhD Data

Mining

P( | )Data

Scientist

Machine

Learning

Expert= 0.52F1:

Features:

P( | , )Data

Scientist

Machine

Learning

Expert= 0.6F3:

5 years

experience

…

Career

path

graph

Impact of Career Path featureAB test with 50:50 split, >10M impressions

CTR

Control group

Group with Career path feature

?

+8%

2%

1%

18

futureme.xing.comfutureme.xing.com

spin-off project which allows for

browsing the career-path graph

19

futureme.xing.com

The professional network

www.xing.com

Thank you@fabianabel

xing.com

futureme.xing.com