SlideShare a Scribd company logo
1 of 16
Download to read offline
/ 16
ISIS 2019
2019.12.06
Development of a Collaborator Recommender System
Based on Directed Graph Model
Kanta Nakamura and Kazushi Okamoto


Department of Informatics, Graduate School of Informatics and Engineering,


The University of Electro-Communications


1
/ 16
2019.12.06 ISIS 2019
Introduction
Importance of utilizing new technology in companies:


• New entry and intensifying international competition


• Shortening life cycle of products and services


• Insuf
fi
cient in-house technology and management resources


Effective use of collaborative research with external organizations


• Selection of researchers who
fi
t purposes


• Researcher recommendation methods


• Collaborator recommender systems
2
/ 16
2019.12.06 ISIS 2019 3
Use of Recommender Systems
When companies use a system directly or


as a University Research Administrators (URA) support


Presented by ranking based on recommendation score
user


(company)
URA
recommender


system
research plan query
recommendation


result
query
recommendation


result
recommendation


result
/ 16
2019.12.06 ISIS 2019
Researcher Recommendation Methods
Network-based recommendation


• Recommendations based on indicators calculated from a link
structure around a researcher pair


• Number and ratio of researchers in common


• Effective for recommendations in the same
fi
eld


Content-based recommendation


• Recommendations based on similarity


of researcher pro
fi
les


• Effective to recommend researchers


across different
fi
elds
4
?
?
Masataka Araki, Marie Katsurai, Ikki Ohmukai, and Hideki Takeda: Interdisciplinary collaborator
recommendation based on research content similarity, IEICE Transactions on Information and
Systems, vol.E100.D, no.4, pp.785-792, 2017.
[araki+, 2017]
/ 16
2019.12.06 ISIS 2019
Researcher Network and Link Prediction
Undirected graph model: researcher contributions are equal


Directed graph model: researcher contributions are not equal


• Ex)
fi
rst author → Nth author, principal researcher → co-researchers
5
: researcher (node)


: known edge


: unknown edge
undirected graph model directed graph model
/ 16
2019.12.06 ISIS 2019
Purpose and Proposed Method
Veri
fi
cation of three directed graph models in researcher recommendation
6
i
<latexit sha1_base64="uxSxj8JaK9/8i56c2Gg8hXw5MC8=">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</latexit>
<latexit sha1_base64="uxSxj8JaK9/8i56c2Gg8hXw5MC8=">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</latexit>
<latexit sha1_base64="uxSxj8JaK9/8i56c2Gg8hXw5MC8=">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</latexit>
<latexit sha1_base64="uxSxj8JaK9/8i56c2Gg8hXw5MC8=">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</latexit>
j
<latexit sha1_base64="D3OmgutMpcRUVq3vuVl8Q345PsM=">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</latexit>
<latexit sha1_base64="D3OmgutMpcRUVq3vuVl8Q345PsM=">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</latexit>
<latexit sha1_base64="D3OmgutMpcRUVq3vuVl8Q345PsM=">AAACZHichVHLSsNAFD2Nr1pf1SIIghRLxVW5EUFxJbpxaa19QC0liVONpklI0oIWf0C3igtXCiLiZ7jxB1z0BwRxWcGNC2/TgKiod5iZM2fuuXNmRrUN3fWImiGpq7unty/cHxkYHBoeiY6O5Vyr5mgiq1mG5RRUxRWGboqsp3uGKNiOUKqqIfLq/mp7P18Xjqtb5qZ3YItSVdkx9YquKR5T6b1yNEEp8iP+E8gBSCCIdSt6gy1sw4KGGqoQMOExNqDA5VaEDILNXAkN5hxGur8vcIQIa2ucJThDYXafxx1eFQPW5HW7puurNT7F4O6wMo4kPdItteiB7uiZ3n+t1fBrtL0c8Kx2tMIujxxPZN7+VVV59rD7qfrTs4cKFn2vOnu3faZ9C62jrx+etzJLG8nGDF3RC/u/pCbd8w3M+qt2nRYbF4jwB8jfn/snyM2lZErJ6fnE8krwFWFMYhqz/N4LWMYa1pHlcwVOcIqz0JM0KMWk8U6qFAo0MXwJaeoD2kyJ6g==</latexit>
<latexit sha1_base64="D3OmgutMpcRUVq3vuVl8Q345PsM=">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</latexit>
principal researcher co-researcher
y 2 {0, 1}
<latexit sha1_base64="DfOOSE2AzLGcz9VuxPXiSKGFPuM=">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</latexit>
<latexit sha1_base64="DfOOSE2AzLGcz9VuxPXiSKGFPuM=">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</latexit>
<latexit sha1_base64="DfOOSE2AzLGcz9VuxPXiSKGFPuM=">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</latexit>
<latexit sha1_base64="DfOOSE2AzLGcz9VuxPXiSKGFPuM=">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</latexit>
node vector
edge vector
Outer product model:


Joint model:


Weighted sum model:
xO = xi ⌦ xj
<latexit sha1_base64="YSh2ovdpQUS1axc+VZL1mv0qOj4=">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</latexit>
xi, xj
<latexit sha1_base64="v7GBeg8Sf0Ha/Mr8RaDtO66iBw4=">AAACjHichVG7SgNBFD1Z3/EVtRFsgiFqIWGiAUURREEsk2hUUAm766hj9sXuJhiX/IBgbWGlICKWtlrZ+AMWfoJYRrCx8GazIBrUu+zcM2fuuXNmRrE04biMPYekpuaW1rb2jnBnV3dPb6Svf80xi7bKc6qpmfaGIjtcEwbPucLV+IZlc1lXNL6uFBZr6+slbjvCNFbdssW3dXnPELtClV2i8pFRb0sxtR2nrFOKHlbynqiMN3AHlXwkxhLMj2gjSAYghiDSZuQKW9iBCRVF6OAw4BLWIMOhbxNJMFjEbcMjziYk/HWOCsKkLVIVpwqZ2AKNezTbDFiD5rWejq9WaReNfpuUUcTZE7tmVfbIbtgL+/i1l+f3qHkpU1bqWm7le48HV97/VemUXex/qf707GIX075XQd4tn6mdQq3rS0en1ZWZbNwbYRfslfyfs2f2QCcwSm/qZYZnzxCmB0j+vO5GsDaRSE4mJjKp2PxC8BTtGMIwxui+pzCPZaSRo31PcIs73Es9UkqalebqpVIo0AzgW0hLnxKnmiU=</latexit>
xW = ↵xi + (1 ↵)xj, ↵ 2 (0, 1)
<latexit sha1_base64="aPNHjVQ1be7hBvC/jDkxHZxSJg4=">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</latexit>
xJ = [xi, xj]
<latexit sha1_base64="IEzFig24OQwpOL+GE5HfZ2Jffrs=">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</latexit>
/ 16
2019.12.06 ISIS 2019
Classi
fi
er and Optimizer
• Output probability


• Interpretation using partial regression coef
fi
cients


• Batch processing can handle large amount of data


• Minimize a log loss function
7
p✓(y = 1|x) = 1
1+exp(✓T ·x)
<latexit sha1_base64="VwaK2vfXLYtO1Fqw1YfFzHb5urk=">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</latexit>
Logistic Regression (LR)
Stochastic Gradient Descent (SGD)


LogLoss = {y log p + (1 y) log(1 p)}
<latexit sha1_base64="NjNb4MpjDGn2BKw7zSXfxbj1oxY=">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</latexit>
/ 16
2019.12.06 ISIS 2019
Overview of the Proposed System
8
outer product


joint


weighted sum
researcher
pro
fi
le set LR
BoW & LDA
node vector edge vector ✓
<latexit sha1_base64="a/AuFkdfRxNSNcbqYURF+q36Umw=">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</latexit>
W
<latexit sha1_base64="XpFfJ/6r8YH8aAQmDZAhQE8XhDE=">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</latexit>
outer product


joint


weighted sum
researcher
pro
fi
le set
LR
BoW & LDA
query
score
training
prediction
/ 16
2019.12.06 ISIS 2019
Grants-in-Aid for Scienti
fi
c Research (KAKEN) DB
9
details of database
publication year FY1964 - FY2018
# of grants 869,081
# of researchers 251,333
characteristics
・Japanese DB

・possible to collect grant awards

・uniquely identi
fi
es researchers

・researcher roles (principal, co-researcher)

・KAKEN API for data collection
As of March 2019
/ 16
2019.12.06 ISIS 2019
Collect Data and Create Researcher Pro
fi
les
10
crawling


(KAKEN API)
morphological
analysis


(MeCab)
researcher
pro
fi
les
grant awards
grant award data


・started in FY1964-FY2017


・title, keyword, abstract


・principal, co-researcher


・researcher number
researcher pro
fi
les


・[researcher number].txt


・nouns only


・space separator
地形 データ 沖 地震


北部 震源 対象 地磁気 地下
水中 震源 以深 ヘリウム


地殻 流体 マントル 起源


流体 震源 筆者 地殻 均質


構造 概念 モデル 構造 均質
弾性 モデル 数値 地域 …
EX) 80421684.txt
/ 16
2019.12.06 ISIS 2019
Generation Samples and Hyper-parameter Tuning
Hyper-parameter tuning


• The number of LDA topics : {50, 100, 150, 200}


• Outer product model : 150


• Joint & weighted sum model : 200


Weighted sum model: α


• α = 0.7 (
fi
xed)
11
starting year (FY) 1964 2012 2013 2014 2015 2017
hyper-parameter
tuning
training edges

4,890,952
test edges

46,046
method evaluation
training edges

5,373,665
test edges

50,512
/ 16
2019.12.06 ISIS 2019
Prediction Accuracy for Each Model
Predict a link based on a probability calculated by logistic regression:
12
model
outer product joint weighted sum
precision 0.553 0.093 0.094
recall 0.693 0.393 0.377
f1 score 0.615 0.150 0.150
yij =
(
1 p(yij = 1|xij) 0.5
0 otherwise.
<latexit sha1_base64="eWUp1tO599q+/FjX8QtitErFsnw=">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</latexit>
/ 16
2019.12.06 ISIS 2019
Computation Time Evaluation for Each Model
• Server performance used in the experiment


• OS: Ubuntu 18.04.1


• CPU model: dual Intel Xeon CPU E5-2697 v3 @ 2.60GHz


• # of cores: 14


• RAM size: 128 GB
13
model [# of edge vector dimensions]
outer product

[22,500]
joint

[400]
weighted sum

[200]
edge vector construction

per edge [s] 

1.87×10−5 1.45×10−6 1.61×10−6
learning [s] 5140 94 49
predicting per edge [s] 1.44×10−6 2.36×10−8 1.14×10−8
/ 16
2019.12.06 ISIS 2019
Analysis of Partial Regression Coef
fi
cient (OP)
• A partial regression coef
fi
cient is high between same topics


• Some pair of different topics have high partial regression coef
fi
cients
14
/ 16
2019.12.06 ISIS 2019
Analysis of Partial Regression Coef
fi
cient (JN & WS)
• Unlike the outer product model, it is dif
fi
cult to
fi
nd characteristics


• Topic co-occurrence between researchers is important for link
prediction
15
joint model weighted sum model
/ 16
2019.12.06 ISIS 2019
Conclusion
Purpose


• Veri
fi
cation of the three directed graph models in researcher
recommendation


Results


• The outer product model achieve highest prediction accuracy


• Topic co-occurrence is important for link prediction


Future work


• Model construction by unsupervised learning


• Veri
fi
cation of data sampling method
16

More Related Content

What's hot

IRJET - Student's Academic Performance Forecasting: Survey
IRJET -  	  Student's Academic Performance Forecasting: SurveyIRJET -  	  Student's Academic Performance Forecasting: Survey
IRJET - Student's Academic Performance Forecasting: SurveyIRJET Journal
 
Assessing the Impact of the Library in the Research Ecosystem: CNI 2018 sprin...
Assessing the Impact of the Library in the Research Ecosystem: CNI 2018 sprin...Assessing the Impact of the Library in the Research Ecosystem: CNI 2018 sprin...
Assessing the Impact of the Library in the Research Ecosystem: CNI 2018 sprin...Megan Hurst
 
Methods for measuring citizen-science impact
Methods for measuring citizen-science impactMethods for measuring citizen-science impact
Methods for measuring citizen-science impactLuigi Ceccaroni
 
Altmetrics, Impact Analysis and Scholarly Communication
Altmetrics, Impact Analysis and Scholarly Communication �Altmetrics, Impact Analysis and Scholarly Communication �
Altmetrics, Impact Analysis and Scholarly Communication Michelle Willmers
 
SCONUL Summer Conference 2018 - Paul Feldman
SCONUL Summer Conference 2018 - Paul FeldmanSCONUL Summer Conference 2018 - Paul Feldman
SCONUL Summer Conference 2018 - Paul Feldmansconul
 
Recommendation of Data Mining Technique in Higher Education Prof. Priya Thaka...
Recommendation of Data Mining Technique in Higher Education Prof. Priya Thaka...Recommendation of Data Mining Technique in Higher Education Prof. Priya Thaka...
Recommendation of Data Mining Technique in Higher Education Prof. Priya Thaka...ijceronline
 
Gabriel Rissola: "Measuring the impact of eInclusion actors"
Gabriel Rissola: "Measuring the impact of eInclusion actors"Gabriel Rissola: "Measuring the impact of eInclusion actors"
Gabriel Rissola: "Measuring the impact of eInclusion actors"TELECENTRE EUROPE
 
The future of scholarly communications professionals
The future of scholarly communications professionalsThe future of scholarly communications professionals
The future of scholarly communications professionalsNancy Pontika
 
Stakeholder engagement in early stage new product-service system development
Stakeholder engagement in early stage new product-service system developmentStakeholder engagement in early stage new product-service system development
Stakeholder engagement in early stage new product-service system developmentMan Hang Yip
 
Can Intra-Organizational Wikis Facilitate Knowledge Transfer and Learning? An...
Can Intra-Organizational Wikis Facilitate Knowledge Transfer and Learning? An...Can Intra-Organizational Wikis Facilitate Knowledge Transfer and Learning? An...
Can Intra-Organizational Wikis Facilitate Knowledge Transfer and Learning? An...Alexander Stocker
 
Prospect for learning analytics to achieve adaptive learning model
Prospect for learning analytics to achieve  adaptive learning modelProspect for learning analytics to achieve  adaptive learning model
Prospect for learning analytics to achieve adaptive learning modelOpen Cyber University of Korea
 
BDVe Webinar Series - Big Data for Public Policy, the state of play - Data-dr...
BDVe Webinar Series - Big Data for Public Policy, the state of play - Data-dr...BDVe Webinar Series - Big Data for Public Policy, the state of play - Data-dr...
BDVe Webinar Series - Big Data for Public Policy, the state of play - Data-dr...Big Data Value Association
 
Developing a PhD Research Topic for Your Research| PhD Assistance UK
Developing a PhD Research Topic for Your Research| PhD Assistance UKDeveloping a PhD Research Topic for Your Research| PhD Assistance UK
Developing a PhD Research Topic for Your Research| PhD Assistance UKPhDAssistanceUK
 
Prospect for learning analytics to achieve adaptive learning model
Prospect for learning analytics to achieve adaptive learning modelProspect for learning analytics to achieve adaptive learning model
Prospect for learning analytics to achieve adaptive learning modelOpen Cyber University of Korea
 
Responsible metrics for research - Jisc Digifest 2016
Responsible metrics for research - Jisc Digifest 2016Responsible metrics for research - Jisc Digifest 2016
Responsible metrics for research - Jisc Digifest 2016Jisc
 
Invited Lecture on Interactive Information Retrieval
Invited Lecture on Interactive Information RetrievalInvited Lecture on Interactive Information Retrieval
Invited Lecture on Interactive Information RetrievalDavidMaxwell77
 
From Data To Information Perspectives On Policy And Practice
From Data To Information  Perspectives On Policy And PracticeFrom Data To Information  Perspectives On Policy And Practice
From Data To Information Perspectives On Policy And PracticeJeff_Watson
 
A New Approach of Analysis of Student Results by using MapReduce
A New Approach of Analysis of Student Results by using MapReduceA New Approach of Analysis of Student Results by using MapReduce
A New Approach of Analysis of Student Results by using MapReduceIRJET Journal
 

What's hot (19)

IRJET - Student's Academic Performance Forecasting: Survey
IRJET -  	  Student's Academic Performance Forecasting: SurveyIRJET -  	  Student's Academic Performance Forecasting: Survey
IRJET - Student's Academic Performance Forecasting: Survey
 
Assessing the Impact of the Library in the Research Ecosystem: CNI 2018 sprin...
Assessing the Impact of the Library in the Research Ecosystem: CNI 2018 sprin...Assessing the Impact of the Library in the Research Ecosystem: CNI 2018 sprin...
Assessing the Impact of the Library in the Research Ecosystem: CNI 2018 sprin...
 
Methods for measuring citizen-science impact
Methods for measuring citizen-science impactMethods for measuring citizen-science impact
Methods for measuring citizen-science impact
 
Altmetrics, Impact Analysis and Scholarly Communication
Altmetrics, Impact Analysis and Scholarly Communication �Altmetrics, Impact Analysis and Scholarly Communication �
Altmetrics, Impact Analysis and Scholarly Communication
 
SCONUL Summer Conference 2018 - Paul Feldman
SCONUL Summer Conference 2018 - Paul FeldmanSCONUL Summer Conference 2018 - Paul Feldman
SCONUL Summer Conference 2018 - Paul Feldman
 
Recommendation of Data Mining Technique in Higher Education Prof. Priya Thaka...
Recommendation of Data Mining Technique in Higher Education Prof. Priya Thaka...Recommendation of Data Mining Technique in Higher Education Prof. Priya Thaka...
Recommendation of Data Mining Technique in Higher Education Prof. Priya Thaka...
 
Assessing and Reporting Research Impact – A Role for the Library - Kristi L....
Assessing and Reporting Research Impact – A Role for the Library  - Kristi L....Assessing and Reporting Research Impact – A Role for the Library  - Kristi L....
Assessing and Reporting Research Impact – A Role for the Library - Kristi L....
 
Gabriel Rissola: "Measuring the impact of eInclusion actors"
Gabriel Rissola: "Measuring the impact of eInclusion actors"Gabriel Rissola: "Measuring the impact of eInclusion actors"
Gabriel Rissola: "Measuring the impact of eInclusion actors"
 
The future of scholarly communications professionals
The future of scholarly communications professionalsThe future of scholarly communications professionals
The future of scholarly communications professionals
 
Stakeholder engagement in early stage new product-service system development
Stakeholder engagement in early stage new product-service system developmentStakeholder engagement in early stage new product-service system development
Stakeholder engagement in early stage new product-service system development
 
Can Intra-Organizational Wikis Facilitate Knowledge Transfer and Learning? An...
Can Intra-Organizational Wikis Facilitate Knowledge Transfer and Learning? An...Can Intra-Organizational Wikis Facilitate Knowledge Transfer and Learning? An...
Can Intra-Organizational Wikis Facilitate Knowledge Transfer and Learning? An...
 
Prospect for learning analytics to achieve adaptive learning model
Prospect for learning analytics to achieve  adaptive learning modelProspect for learning analytics to achieve  adaptive learning model
Prospect for learning analytics to achieve adaptive learning model
 
BDVe Webinar Series - Big Data for Public Policy, the state of play - Data-dr...
BDVe Webinar Series - Big Data for Public Policy, the state of play - Data-dr...BDVe Webinar Series - Big Data for Public Policy, the state of play - Data-dr...
BDVe Webinar Series - Big Data for Public Policy, the state of play - Data-dr...
 
Developing a PhD Research Topic for Your Research| PhD Assistance UK
Developing a PhD Research Topic for Your Research| PhD Assistance UKDeveloping a PhD Research Topic for Your Research| PhD Assistance UK
Developing a PhD Research Topic for Your Research| PhD Assistance UK
 
Prospect for learning analytics to achieve adaptive learning model
Prospect for learning analytics to achieve adaptive learning modelProspect for learning analytics to achieve adaptive learning model
Prospect for learning analytics to achieve adaptive learning model
 
Responsible metrics for research - Jisc Digifest 2016
Responsible metrics for research - Jisc Digifest 2016Responsible metrics for research - Jisc Digifest 2016
Responsible metrics for research - Jisc Digifest 2016
 
Invited Lecture on Interactive Information Retrieval
Invited Lecture on Interactive Information RetrievalInvited Lecture on Interactive Information Retrieval
Invited Lecture on Interactive Information Retrieval
 
From Data To Information Perspectives On Policy And Practice
From Data To Information  Perspectives On Policy And PracticeFrom Data To Information  Perspectives On Policy And Practice
From Data To Information Perspectives On Policy And Practice
 
A New Approach of Analysis of Student Results by using MapReduce
A New Approach of Analysis of Student Results by using MapReduceA New Approach of Analysis of Student Results by using MapReduce
A New Approach of Analysis of Student Results by using MapReduce
 

Similar to Development of a Collaborator Recommender System Based on Directed Graph Model

A Review On Recommender Systems For University Admissions
A Review On Recommender Systems For University AdmissionsA Review On Recommender Systems For University Admissions
A Review On Recommender Systems For University AdmissionsBecky Gilbert
 
University Recommendation Support System using ML Algorithms
University Recommendation Support System using ML AlgorithmsUniversity Recommendation Support System using ML Algorithms
University Recommendation Support System using ML AlgorithmsIRJET Journal
 
Survey on Study Partners Recommendation for Online Courses
Survey on Study Partners Recommendation for Online CoursesSurvey on Study Partners Recommendation for Online Courses
Survey on Study Partners Recommendation for Online CoursesIRJET Journal
 
SAS Slides v15
SAS Slides v15SAS Slides v15
SAS Slides v15Ron Wen
 
Tutorial on Bias in Rec Sys @ UMAP2020
Tutorial on Bias in Rec Sys @ UMAP2020Tutorial on Bias in Rec Sys @ UMAP2020
Tutorial on Bias in Rec Sys @ UMAP2020Mirko Marras
 
CORE Analytics Dashboard
CORE Analytics DashboardCORE Analytics Dashboard
CORE Analytics Dashboardpetrknoth
 
A Comprehensive Review of Relevant Techniques used in Course Recommendation S...
A Comprehensive Review of Relevant Techniques used in Course Recommendation S...A Comprehensive Review of Relevant Techniques used in Course Recommendation S...
A Comprehensive Review of Relevant Techniques used in Course Recommendation S...IRJET Journal
 
Transforming How Sponsors and CROs Interact with Clinical Sites
Transforming How Sponsors and CROs Interact with Clinical SitesTransforming How Sponsors and CROs Interact with Clinical Sites
Transforming How Sponsors and CROs Interact with Clinical SitesPerficient, Inc.
 
Introduction to Recommendation Systems
Introduction to Recommendation SystemsIntroduction to Recommendation Systems
Introduction to Recommendation SystemsZia Babar
 
Motivating Academic Research at Undergraduate Level: Participation in the Sci...
Motivating Academic Research at Undergraduate Level: Participation in the Sci...Motivating Academic Research at Undergraduate Level: Participation in the Sci...
Motivating Academic Research at Undergraduate Level: Participation in the Sci...mohamis
 
A Systematic Literature Survey On Recommendation System
A Systematic Literature Survey On Recommendation SystemA Systematic Literature Survey On Recommendation System
A Systematic Literature Survey On Recommendation SystemGina Rizzo
 
Asareca research management guidelines
Asareca research management guidelinesAsareca research management guidelines
Asareca research management guidelinesFrancois Stepman
 
DATA ANALYTICS FOR HIGHER EDUCATION
 DATA ANALYTICS FOR HIGHER EDUCATION DATA ANALYTICS FOR HIGHER EDUCATION
DATA ANALYTICS FOR HIGHER EDUCATIONSamantha Suraweera
 
IRJET- A Survey on Recommender Systems used for User Service Rating in Social...
IRJET- A Survey on Recommender Systems used for User Service Rating in Social...IRJET- A Survey on Recommender Systems used for User Service Rating in Social...
IRJET- A Survey on Recommender Systems used for User Service Rating in Social...IRJET Journal
 
Classification of Researcher's Collaboration Patterns Towards Research Perfor...
Classification of Researcher's Collaboration Patterns Towards Research Perfor...Classification of Researcher's Collaboration Patterns Towards Research Perfor...
Classification of Researcher's Collaboration Patterns Towards Research Perfor...Nur Hazimah Khalid
 
Multicriteria and cost benefit analysis for smart grid projects
Multicriteria and cost benefit analysis for smart grid projectsMulticriteria and cost benefit analysis for smart grid projects
Multicriteria and cost benefit analysis for smart grid projectsLeonardo ENERGY
 
Final business competition presentation (1)
Final business competition presentation (1)Final business competition presentation (1)
Final business competition presentation (1)Santhosh Prasad
 
IRJET - Recommendation of Branch of Engineering using Machine Learning
IRJET - Recommendation of Branch of Engineering using Machine LearningIRJET - Recommendation of Branch of Engineering using Machine Learning
IRJET - Recommendation of Branch of Engineering using Machine LearningIRJET Journal
 

Similar to Development of a Collaborator Recommender System Based on Directed Graph Model (20)

A Review On Recommender Systems For University Admissions
A Review On Recommender Systems For University AdmissionsA Review On Recommender Systems For University Admissions
A Review On Recommender Systems For University Admissions
 
University Recommendation Support System using ML Algorithms
University Recommendation Support System using ML AlgorithmsUniversity Recommendation Support System using ML Algorithms
University Recommendation Support System using ML Algorithms
 
Survey on Study Partners Recommendation for Online Courses
Survey on Study Partners Recommendation for Online CoursesSurvey on Study Partners Recommendation for Online Courses
Survey on Study Partners Recommendation for Online Courses
 
SAS Slides v15
SAS Slides v15SAS Slides v15
SAS Slides v15
 
Tutorial on Bias in Rec Sys @ UMAP2020
Tutorial on Bias in Rec Sys @ UMAP2020Tutorial on Bias in Rec Sys @ UMAP2020
Tutorial on Bias in Rec Sys @ UMAP2020
 
Carpenter, "RA21 Update"
Carpenter, "RA21 Update"Carpenter, "RA21 Update"
Carpenter, "RA21 Update"
 
CORE Analytics Dashboard
CORE Analytics DashboardCORE Analytics Dashboard
CORE Analytics Dashboard
 
A Comprehensive Review of Relevant Techniques used in Course Recommendation S...
A Comprehensive Review of Relevant Techniques used in Course Recommendation S...A Comprehensive Review of Relevant Techniques used in Course Recommendation S...
A Comprehensive Review of Relevant Techniques used in Course Recommendation S...
 
Transforming How Sponsors and CROs Interact with Clinical Sites
Transforming How Sponsors and CROs Interact with Clinical SitesTransforming How Sponsors and CROs Interact with Clinical Sites
Transforming How Sponsors and CROs Interact with Clinical Sites
 
Introduction to Recommendation Systems
Introduction to Recommendation SystemsIntroduction to Recommendation Systems
Introduction to Recommendation Systems
 
Motivating Academic Research at Undergraduate Level: Participation in the Sci...
Motivating Academic Research at Undergraduate Level: Participation in the Sci...Motivating Academic Research at Undergraduate Level: Participation in the Sci...
Motivating Academic Research at Undergraduate Level: Participation in the Sci...
 
A Systematic Literature Survey On Recommendation System
A Systematic Literature Survey On Recommendation SystemA Systematic Literature Survey On Recommendation System
A Systematic Literature Survey On Recommendation System
 
Asareca research management guidelines
Asareca research management guidelinesAsareca research management guidelines
Asareca research management guidelines
 
DATA ANALYTICS FOR HIGHER EDUCATION
 DATA ANALYTICS FOR HIGHER EDUCATION DATA ANALYTICS FOR HIGHER EDUCATION
DATA ANALYTICS FOR HIGHER EDUCATION
 
IRJET- A Survey on Recommender Systems used for User Service Rating in Social...
IRJET- A Survey on Recommender Systems used for User Service Rating in Social...IRJET- A Survey on Recommender Systems used for User Service Rating in Social...
IRJET- A Survey on Recommender Systems used for User Service Rating in Social...
 
Paper publication
Paper publicationPaper publication
Paper publication
 
Classification of Researcher's Collaboration Patterns Towards Research Perfor...
Classification of Researcher's Collaboration Patterns Towards Research Perfor...Classification of Researcher's Collaboration Patterns Towards Research Perfor...
Classification of Researcher's Collaboration Patterns Towards Research Perfor...
 
Multicriteria and cost benefit analysis for smart grid projects
Multicriteria and cost benefit analysis for smart grid projectsMulticriteria and cost benefit analysis for smart grid projects
Multicriteria and cost benefit analysis for smart grid projects
 
Final business competition presentation (1)
Final business competition presentation (1)Final business competition presentation (1)
Final business competition presentation (1)
 
IRJET - Recommendation of Branch of Engineering using Machine Learning
IRJET - Recommendation of Branch of Engineering using Machine LearningIRJET - Recommendation of Branch of Engineering using Machine Learning
IRJET - Recommendation of Branch of Engineering using Machine Learning
 

More from Okamoto Laboratory, The University of Electro-Communications

More from Okamoto Laboratory, The University of Electro-Communications (18)

クラウドソーシングにおける協調的な共同作業に対する組織構成システム
クラウドソーシングにおける協調的な共同作業に対する組織構成システムクラウドソーシングにおける協調的な共同作業に対する組織構成システム
クラウドソーシングにおける協調的な共同作業に対する組織構成システム
 
リンク予測に基づく共同研究者推薦システムの試作
リンク予測に基づく共同研究者推薦システムの試作リンク予測に基づく共同研究者推薦システムの試作
リンク予測に基づく共同研究者推薦システムの試作
 
Visualizing the Importance of Floor-Plan Image Features in Rent-Prediction Mo...
Visualizing the Importance of Floor-Plan Image Features in Rent-Prediction Mo...Visualizing the Importance of Floor-Plan Image Features in Rent-Prediction Mo...
Visualizing the Importance of Floor-Plan Image Features in Rent-Prediction Mo...
 
間取り図を用いた賃料予測モデルに関する一検討
間取り図を用いた賃料予測モデルに関する一検討間取り図を用いた賃料予測モデルに関する一検討
間取り図を用いた賃料予測モデルに関する一検討
 
Rent Prediction Models with Floor Plan Images
Rent Prediction Models with Floor Plan ImagesRent Prediction Models with Floor Plan Images
Rent Prediction Models with Floor Plan Images
 
分散表現を用いたリアルタイム学習型セッションベース推薦システム
分散表現を用いたリアルタイム学習型セッションベース推薦システム分散表現を用いたリアルタイム学習型セッションベース推薦システム
分散表現を用いたリアルタイム学習型セッションベース推薦システム
 
アイテム分散表現の階層化・集約演算に基づくセッションベース推薦システム
アイテム分散表現の階層化・集約演算に基づくセッションベース推薦システムアイテム分散表現の階層化・集約演算に基づくセッションベース推薦システム
アイテム分散表現の階層化・集約演算に基づくセッションベース推薦システム
 
発売日前のレビューとPU-Learningを用いた
スパムレビュー検出
発売日前のレビューとPU-Learningを用いた
スパムレビュー検出発売日前のレビューとPU-Learningを用いた
スパムレビュー検出
発売日前のレビューとPU-Learningを用いた
スパムレビュー検出
 
モデルベース協調フィルタリングにおける推薦の透明性に関する検討
モデルベース協調フィルタリングにおける推薦の透明性に関する検討モデルベース協調フィルタリングにおける推薦の透明性に関する検討
モデルベース協調フィルタリングにおける推薦の透明性に関する検討
 
重回帰分析による推薦の透明性を有したモデルベース協調フィルタリング
重回帰分析による推薦の透明性を有したモデルベース協調フィルタリング重回帰分析による推薦の透明性を有したモデルベース協調フィルタリング
重回帰分析による推薦の透明性を有したモデルベース協調フィルタリング
 
Word2Vecによる次元圧縮と重回帰分析型協調フィルタリングへの応用
Word2Vecによる次元圧縮と重回帰分析型協調フィルタリングへの応用Word2Vecによる次元圧縮と重回帰分析型協調フィルタリングへの応用
Word2Vecによる次元圧縮と重回帰分析型協調フィルタリングへの応用
 
数式からみるWord2Vec
数式からみるWord2Vec数式からみるWord2Vec
数式からみるWord2Vec
 
Rによるベイジアンネットワーク入門
Rによるベイジアンネットワーク入門Rによるベイジアンネットワーク入門
Rによるベイジアンネットワーク入門
 
単語の分散表現の 購買履歴への応用
単語の分散表現の 購買履歴への応用単語の分散表現の 購買履歴への応用
単語の分散表現の 購買履歴への応用
 
機関リポジトリから収集した学術論文のテキスト解析に関する一検討
機関リポジトリから収集した学術論文のテキスト解析に関する一検討機関リポジトリから収集した学術論文のテキスト解析に関する一検討
機関リポジトリから収集した学術論文のテキスト解析に関する一検討
 
Text Analysis of Academic Papers Archived in Institutional Repositories
Text Analysis of Academic Papers Archived in Institutional RepositoriesText Analysis of Academic Papers Archived in Institutional Repositories
Text Analysis of Academic Papers Archived in Institutional Repositories
 
Families of Triangular Norm Based Kernel Function and Its Application to Kern...
Families of Triangular Norm Based Kernel Function and Its Application to Kern...Families of Triangular Norm Based Kernel Function and Its Application to Kern...
Families of Triangular Norm Based Kernel Function and Its Application to Kern...
 
これから始めるディープラーニング
これから始めるディープラーニングこれから始めるディープラーニング
これから始めるディープラーニング
 

Recently uploaded

Dubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls DubaiDubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls Dubaihf8803863
 
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /WhatsappsBeautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsappssapnasaifi408
 
Call Girls In Dwarka 9654467111 Escorts Service
Call Girls In Dwarka 9654467111 Escorts ServiceCall Girls In Dwarka 9654467111 Escorts Service
Call Girls In Dwarka 9654467111 Escorts ServiceSapana Sha
 
Industrialised data - the key to AI success.pdf
Industrialised data - the key to AI success.pdfIndustrialised data - the key to AI success.pdf
Industrialised data - the key to AI success.pdfLars Albertsson
 
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)jennyeacort
 
9654467111 Call Girls In Munirka Hotel And Home Service
9654467111 Call Girls In Munirka Hotel And Home Service9654467111 Call Girls In Munirka Hotel And Home Service
9654467111 Call Girls In Munirka Hotel And Home ServiceSapana Sha
 
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样vhwb25kk
 
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024thyngster
 
ASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel CanterASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel Cantervoginip
 
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...Suhani Kapoor
 
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130Suhani Kapoor
 
PKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptxPKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptxPramod Kumar Srivastava
 
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort servicejennyeacort
 
How we prevented account sharing with MFA
How we prevented account sharing with MFAHow we prevented account sharing with MFA
How we prevented account sharing with MFAAndrei Kaleshka
 
DBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfDBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfJohn Sterrett
 
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...Sapana Sha
 
20240419 - Measurecamp Amsterdam - SAM.pdf
20240419 - Measurecamp Amsterdam - SAM.pdf20240419 - Measurecamp Amsterdam - SAM.pdf
20240419 - Measurecamp Amsterdam - SAM.pdfHuman37
 
Customer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptxCustomer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptxEmmanuel Dauda
 
毕业文凭制作#回国入职#diploma#degree澳洲中央昆士兰大学毕业证成绩单pdf电子版制作修改#毕业文凭制作#回国入职#diploma#degree
毕业文凭制作#回国入职#diploma#degree澳洲中央昆士兰大学毕业证成绩单pdf电子版制作修改#毕业文凭制作#回国入职#diploma#degree毕业文凭制作#回国入职#diploma#degree澳洲中央昆士兰大学毕业证成绩单pdf电子版制作修改#毕业文凭制作#回国入职#diploma#degree
毕业文凭制作#回国入职#diploma#degree澳洲中央昆士兰大学毕业证成绩单pdf电子版制作修改#毕业文凭制作#回国入职#diploma#degreeyuu sss
 

Recently uploaded (20)

Dubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls DubaiDubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
 
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /WhatsappsBeautiful Sapna Vip  Call Girls Hauz Khas 9711199012 Call /Whatsapps
Beautiful Sapna Vip Call Girls Hauz Khas 9711199012 Call /Whatsapps
 
Deep Generative Learning for All - The Gen AI Hype (Spring 2024)
Deep Generative Learning for All - The Gen AI Hype (Spring 2024)Deep Generative Learning for All - The Gen AI Hype (Spring 2024)
Deep Generative Learning for All - The Gen AI Hype (Spring 2024)
 
Call Girls In Dwarka 9654467111 Escorts Service
Call Girls In Dwarka 9654467111 Escorts ServiceCall Girls In Dwarka 9654467111 Escorts Service
Call Girls In Dwarka 9654467111 Escorts Service
 
Industrialised data - the key to AI success.pdf
Industrialised data - the key to AI success.pdfIndustrialised data - the key to AI success.pdf
Industrialised data - the key to AI success.pdf
 
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
 
9654467111 Call Girls In Munirka Hotel And Home Service
9654467111 Call Girls In Munirka Hotel And Home Service9654467111 Call Girls In Munirka Hotel And Home Service
9654467111 Call Girls In Munirka Hotel And Home Service
 
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样
1:1定制(UQ毕业证)昆士兰大学毕业证成绩单修改留信学历认证原版一模一样
 
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
Consent & Privacy Signals on Google *Pixels* - MeasureCamp Amsterdam 2024
 
ASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel CanterASML's Taxonomy Adventure by Daniel Canter
ASML's Taxonomy Adventure by Daniel Canter
 
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
VIP High Class Call Girls Jamshedpur Anushka 8250192130 Independent Escort Se...
 
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
 
PKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptxPKS-TGC-1084-630 - Stage 1 Proposal.pptx
PKS-TGC-1084-630 - Stage 1 Proposal.pptx
 
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
 
How we prevented account sharing with MFA
How we prevented account sharing with MFAHow we prevented account sharing with MFA
How we prevented account sharing with MFA
 
DBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdfDBA Basics: Getting Started with Performance Tuning.pdf
DBA Basics: Getting Started with Performance Tuning.pdf
 
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...
Saket, (-DELHI )+91-9654467111-(=)CHEAP Call Girls in Escorts Service Saket C...
 
20240419 - Measurecamp Amsterdam - SAM.pdf
20240419 - Measurecamp Amsterdam - SAM.pdf20240419 - Measurecamp Amsterdam - SAM.pdf
20240419 - Measurecamp Amsterdam - SAM.pdf
 
Customer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptxCustomer Service Analytics - Make Sense of All Your Data.pptx
Customer Service Analytics - Make Sense of All Your Data.pptx
 
毕业文凭制作#回国入职#diploma#degree澳洲中央昆士兰大学毕业证成绩单pdf电子版制作修改#毕业文凭制作#回国入职#diploma#degree
毕业文凭制作#回国入职#diploma#degree澳洲中央昆士兰大学毕业证成绩单pdf电子版制作修改#毕业文凭制作#回国入职#diploma#degree毕业文凭制作#回国入职#diploma#degree澳洲中央昆士兰大学毕业证成绩单pdf电子版制作修改#毕业文凭制作#回国入职#diploma#degree
毕业文凭制作#回国入职#diploma#degree澳洲中央昆士兰大学毕业证成绩单pdf电子版制作修改#毕业文凭制作#回国入职#diploma#degree
 

Development of a Collaborator Recommender System Based on Directed Graph Model

  • 1. / 16 ISIS 2019 2019.12.06 Development of a Collaborator Recommender System Based on Directed Graph Model Kanta Nakamura and Kazushi Okamoto Department of Informatics, Graduate School of Informatics and Engineering, The University of Electro-Communications 1
  • 2. / 16 2019.12.06 ISIS 2019 Introduction Importance of utilizing new technology in companies: • New entry and intensifying international competition • Shortening life cycle of products and services • Insuf fi cient in-house technology and management resources Effective use of collaborative research with external organizations • Selection of researchers who fi t purposes • Researcher recommendation methods • Collaborator recommender systems 2
  • 3. / 16 2019.12.06 ISIS 2019 3 Use of Recommender Systems When companies use a system directly or 
 as a University Research Administrators (URA) support Presented by ranking based on recommendation score user (company) URA recommender system research plan query recommendation result query recommendation result recommendation result
  • 4. / 16 2019.12.06 ISIS 2019 Researcher Recommendation Methods Network-based recommendation • Recommendations based on indicators calculated from a link structure around a researcher pair • Number and ratio of researchers in common • Effective for recommendations in the same fi eld Content-based recommendation • Recommendations based on similarity 
 of researcher pro fi les • Effective to recommend researchers 
 across different fi elds 4 ? ? Masataka Araki, Marie Katsurai, Ikki Ohmukai, and Hideki Takeda: Interdisciplinary collaborator recommendation based on research content similarity, IEICE Transactions on Information and Systems, vol.E100.D, no.4, pp.785-792, 2017. [araki+, 2017]
  • 5. / 16 2019.12.06 ISIS 2019 Researcher Network and Link Prediction Undirected graph model: researcher contributions are equal Directed graph model: researcher contributions are not equal • Ex) fi rst author → Nth author, principal researcher → co-researchers 5 : researcher (node) : known edge : unknown edge undirected graph model directed graph model
  • 6. / 16 2019.12.06 ISIS 2019 Purpose and Proposed Method Veri fi cation of three directed graph models in researcher recommendation 6 i <latexit sha1_base64="uxSxj8JaK9/8i56c2Gg8hXw5MC8=">AAACZHichVHLSsNAFD2Nr1qrrYogCCIWxVW5EUFxVXTj0rZWhVpKEkcdmiYhSQu1+AO6VVy4UhARP8ONP+DCHxDEpYIbF96mAVFR7zAzZ87cc+fMjO6Y0vOJHiJKR2dXd0+0N9YX7x9IJAeH1j275hqiYNim7W7qmidMaYmCL31TbDqu0Kq6KTb0ynJrf6MuXE/a1prfcESpqu1ackcams9UVpaTKUpTEBM/gRqCFMJYtZNX2MI2bBiooQoBCz5jExo8bkWoIDjMldBkzmUkg32BA8RYW+MswRkasxUed3lVDFmL162aXqA2+BSTu8vKCUzRPV3TC93RDT3R+6+1mkGNlpcGz3pbK5xy4nA0//avqsqzj71P1Z+efexgIfAq2bsTMK1bGG19ff/0Jb+Ym2pO0wU9s/9zeqBbvoFVfzUusyJ3hhh/gPr9uX+C9dm0Smk1O5fKLIVfEcUYJjHD7z2PDFawigKfK3CEY5xEHpW4MqyMtFOVSKgZxpdQxj8A2EyJ6Q==</latexit> <latexit sha1_base64="uxSxj8JaK9/8i56c2Gg8hXw5MC8=">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</latexit> <latexit sha1_base64="uxSxj8JaK9/8i56c2Gg8hXw5MC8=">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</latexit> <latexit sha1_base64="uxSxj8JaK9/8i56c2Gg8hXw5MC8=">AAACZHichVHLSsNAFD2Nr1qrrYogCCIWxVW5EUFxVXTj0rZWhVpKEkcdmiYhSQu1+AO6VVy4UhARP8ONP+DCHxDEpYIbF96mAVFR7zAzZ87cc+fMjO6Y0vOJHiJKR2dXd0+0N9YX7x9IJAeH1j275hqiYNim7W7qmidMaYmCL31TbDqu0Kq6KTb0ynJrf6MuXE/a1prfcESpqu1ackcams9UVpaTKUpTEBM/gRqCFMJYtZNX2MI2bBiooQoBCz5jExo8bkWoIDjMldBkzmUkg32BA8RYW+MswRkasxUed3lVDFmL162aXqA2+BSTu8vKCUzRPV3TC93RDT3R+6+1mkGNlpcGz3pbK5xy4nA0//avqsqzj71P1Z+efexgIfAq2bsTMK1bGG19ff/0Jb+Ym2pO0wU9s/9zeqBbvoFVfzUusyJ3hhh/gPr9uX+C9dm0Smk1O5fKLIVfEcUYJjHD7z2PDFawigKfK3CEY5xEHpW4MqyMtFOVSKgZxpdQxj8A2EyJ6Q==</latexit> j <latexit sha1_base64="D3OmgutMpcRUVq3vuVl8Q345PsM=">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</latexit> <latexit sha1_base64="D3OmgutMpcRUVq3vuVl8Q345PsM=">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</latexit> <latexit sha1_base64="D3OmgutMpcRUVq3vuVl8Q345PsM=">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</latexit> <latexit sha1_base64="D3OmgutMpcRUVq3vuVl8Q345PsM=">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</latexit> principal researcher co-researcher y 2 {0, 1} <latexit sha1_base64="DfOOSE2AzLGcz9VuxPXiSKGFPuM=">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</latexit> <latexit sha1_base64="DfOOSE2AzLGcz9VuxPXiSKGFPuM=">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</latexit> <latexit sha1_base64="DfOOSE2AzLGcz9VuxPXiSKGFPuM=">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</latexit> <latexit sha1_base64="DfOOSE2AzLGcz9VuxPXiSKGFPuM=">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</latexit> node vector edge vector Outer product model: Joint model: Weighted sum model: xO = xi ⌦ xj <latexit sha1_base64="YSh2ovdpQUS1axc+VZL1mv0qOj4=">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</latexit> xi, xj <latexit sha1_base64="v7GBeg8Sf0Ha/Mr8RaDtO66iBw4=">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</latexit> xW = ↵xi + (1 ↵)xj, ↵ 2 (0, 1) <latexit sha1_base64="aPNHjVQ1be7hBvC/jDkxHZxSJg4=">AAACxXichVHLShxBFD12XmbycBI3ATeNg2EkOtw2giIExCySpY+MIzgy6W5LrVj9oLtmcNI07v0BF1kZEAn5DEH8ARd+giQ7E91kkTs9DSEZktyi6t46555bt6qcUMlYE533GTdu3rp9p/9u4d79Bw8Hio8eL8dBM3JF1Q1UEK04diyU9EVVS63EShgJ23OUqDnbLzt8rSWiWAb+G90OxZpnb/pyQ7q2ZqhRfJvUnUCtx22PnbmTNpJaar4w67YKt+weTqbmM7NsjXfp0R7+XTqWS+vSN8s0Zo02iiWqUGZmb2DlQQm5zQfFI9SxjgAumvAg4ENzrGAj5rEKC4SQsTUkjEUcyYwXSFFgbZOzBGfYjG7zusm71Rz1ed+pGWdql09RPCNWmhihM/pEl3RKn+mCfvy1VpLV6PTSZu90tSJsDOw9Wbr+r8pjr7H1S/XPnjU2MJ31Krn3MEM6t3C7+tb7/culmcWR5Cl9pC/c/wGd0zHfwG99cw8XxOIHFPgDrD+fuzdYnqhYzysTC5Ol2bn8K/oxhGGU+b2nMIvXmEeVzz3BV3zHlfHK8AxttLqpRl+uGcRvZuz+BNferz4=</latexit> xJ = [xi, xj] <latexit sha1_base64="IEzFig24OQwpOL+GE5HfZ2Jffrs=">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</latexit>
  • 7. / 16 2019.12.06 ISIS 2019 Classi fi er and Optimizer • Output probability • Interpretation using partial regression coef fi cients • Batch processing can handle large amount of data • Minimize a log loss function 7 p✓(y = 1|x) = 1 1+exp(✓T ·x) <latexit sha1_base64="VwaK2vfXLYtO1Fqw1YfFzHb5urk=">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</latexit> Logistic Regression (LR) Stochastic Gradient Descent (SGD) LogLoss = {y log p + (1 y) log(1 p)} <latexit sha1_base64="NjNb4MpjDGn2BKw7zSXfxbj1oxY=">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</latexit>
  • 8. / 16 2019.12.06 ISIS 2019 Overview of the Proposed System 8 outer product joint weighted sum researcher pro fi le set LR BoW & LDA node vector edge vector ✓ <latexit sha1_base64="a/AuFkdfRxNSNcbqYURF+q36Umw=">AAACdnichVFNLwNRFD0dX1UfLTYSiYimWDW3SIhVw8YSVSQqzcx4dNL5ysxrpRp/wB+wEAsSRPwMG3/Aoj9BLElYWLidTiII7mTePe+8e+477z3NNQ1fEjUiSlt7R2dXtDvW09vXH08MDG74TsXTRV53TMfb0lRfmIYt8tKQpthyPaFamik2tfJSc32zKjzfcOx1WXPFjqXu28aeoauSqWJioKA55q5fszgVZElItZhIUpqCGPsJMiFIIowVJ3GNAnbhQEcFFgRsSMYmVPj8bSMDgsvcDurMeYyMYF3gCDHWVrhKcIXKbJnHfZ5th6zN82ZPP1DrvIvJv8fKMaTogW7ome7plh7p/dde9aBH00uNs9bSCrcYPx7Ovf6rsjhLlD5Vf3qW2MN84NVg727ANE+ht/TVw5Pn3MJaqj5BF/TE/s+pQXd8Arv6ol+uirVTxPgBMt+v+yfYmE5nZtLTq7PJ7GL4FFGMYBxTfN9zyGIZK8jzvgc4wxWuI2/KqJJSJlulSiTUDOFLKPQBKG+RWg==</latexit> W <latexit sha1_base64="XpFfJ/6r8YH8aAQmDZAhQE8XhDE=">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</latexit> outer product joint weighted sum researcher pro fi le set LR BoW & LDA query score training prediction
  • 9. / 16 2019.12.06 ISIS 2019 Grants-in-Aid for Scienti fi c Research (KAKEN) DB 9 details of database publication year FY1964 - FY2018 # of grants 869,081 # of researchers 251,333 characteristics ・Japanese DB ・possible to collect grant awards ・uniquely identi fi es researchers ・researcher roles (principal, co-researcher) ・KAKEN API for data collection As of March 2019
  • 10. / 16 2019.12.06 ISIS 2019 Collect Data and Create Researcher Pro fi les 10 crawling (KAKEN API) morphological analysis (MeCab) researcher pro fi les grant awards grant award data ・started in FY1964-FY2017 ・title, keyword, abstract ・principal, co-researcher ・researcher number researcher pro fi les ・[researcher number].txt ・nouns only ・space separator 地形 データ 沖 地震 北部 震源 対象 地磁気 地下 水中 震源 以深 ヘリウム 地殻 流体 マントル 起源 流体 震源 筆者 地殻 均質 構造 概念 モデル 構造 均質 弾性 モデル 数値 地域 … EX) 80421684.txt
  • 11. / 16 2019.12.06 ISIS 2019 Generation Samples and Hyper-parameter Tuning Hyper-parameter tuning • The number of LDA topics : {50, 100, 150, 200} • Outer product model : 150 • Joint & weighted sum model : 200 Weighted sum model: α • α = 0.7 ( fi xed) 11 starting year (FY) 1964 2012 2013 2014 2015 2017 hyper-parameter tuning training edges 4,890,952 test edges 46,046 method evaluation training edges 5,373,665 test edges 50,512
  • 12. / 16 2019.12.06 ISIS 2019 Prediction Accuracy for Each Model Predict a link based on a probability calculated by logistic regression: 12 model outer product joint weighted sum precision 0.553 0.093 0.094 recall 0.693 0.393 0.377 f1 score 0.615 0.150 0.150 yij = ( 1 p(yij = 1|xij) 0.5 0 otherwise. <latexit sha1_base64="eWUp1tO599q+/FjX8QtitErFsnw=">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</latexit>
  • 13. / 16 2019.12.06 ISIS 2019 Computation Time Evaluation for Each Model • Server performance used in the experiment • OS: Ubuntu 18.04.1 • CPU model: dual Intel Xeon CPU E5-2697 v3 @ 2.60GHz • # of cores: 14 • RAM size: 128 GB 13 model [# of edge vector dimensions] outer product [22,500] joint [400] weighted sum [200] edge vector construction per edge [s] 1.87×10−5 1.45×10−6 1.61×10−6 learning [s] 5140 94 49 predicting per edge [s] 1.44×10−6 2.36×10−8 1.14×10−8
  • 14. / 16 2019.12.06 ISIS 2019 Analysis of Partial Regression Coef fi cient (OP) • A partial regression coef fi cient is high between same topics • Some pair of different topics have high partial regression coef fi cients 14
  • 15. / 16 2019.12.06 ISIS 2019 Analysis of Partial Regression Coef fi cient (JN & WS) • Unlike the outer product model, it is dif fi cult to fi nd characteristics • Topic co-occurrence between researchers is important for link prediction 15 joint model weighted sum model
  • 16. / 16 2019.12.06 ISIS 2019 Conclusion Purpose • Veri fi cation of the three directed graph models in researcher recommendation Results • The outer product model achieve highest prediction accuracy • Topic co-occurrence is important for link prediction Future work • Model construction by unsupervised learning • Veri fi cation of data sampling method 16