CORRELATING R & D EXPENDITURE AND SCHOLARLY PUBLICATION OUTPUT USING K-MEANS CLUSTERING

International Journal of Information Technology, Modeling and Computing (IJITMC) Vol. 5, No.1, February 2017
DOI: 10.5121/ijitmc.2017.5101 1
CORRELATING R & D EXPENDITURE AND SCHOLARLY
PUBLICATION OUTPUT USING K-MEANS CLUSTERING
R. S. Kamath 1
, R. K. Kamat 2
and S.M. Pujar 3
1
Department of Computer Studies, Chhatrapati Shahu Institute of Business Education
and Research, Kolhapur, India
2
Department of Electronics, Shivaji University, Kolhapur, India
3
Indira Gandhi Institute of Development Research, Gen AK Vaidya Marg, Goregaon
East, Mumbai, India
ABSTRACT
History of humanity especially post renaissance era depicts the contribution of research and its output In
terms of publication, patents and technology transfer paving the way for the societal prosperity. Scientific
writing and research publication are fundamental components indicative of academic excellence and
supposedly committed to the Research and Development (R&D) funding. Thus the budget allocation and
expenditure thereof towards R&D is considered to be a vital parameter for the advancement in science and
technology and also for social and economic well being. In view of the tall aspirations of society and
government at large it becomes indispensable to investigate whether the scholastic output is going hand in
hand with the R & D budget spillover or otherwise. We present in this communication a systematic
clustering approach based on K-means algorithm to reveal the impact of R&D expenditure on the extent of
research publications. Two independent sources of data, Research and development expenditure i.e.
percentage of Gross Domestic Product (GDP) and Scientific and technical journal articles, are brought
together in this comprehensive study. From an empirical perspective, present study found that there exist a
positive linear correlation between R & D Expenditure and number of research publication.
KEYWORDS
R&D Expenditure, Publications, Correlation, Clustering, Dataset, Analysis
1. INTRODUCTION
The following meaningful quote by astronomer Carl Sagan “Somewhere, something incredible is
waiting to be known,” essentially portrays the psyche of a passionate researcher, how however
fails to perform in the crunch of diminishing R & D budgetary allocation. It is without any doubt
that the investment towards R&D influences innovations and in turn stimulates the growth of a
country [11]. However the relationship between the percentage of GDP spent on the R&D vis-à-
vis the quantity as well as quantity of scholarly journal articles published has been a topic of
inquiry through the globe. There are number of studies reported in the past investigating the
correlation of productivity in the academic research community in the light of the research
expenditure [1, 2, 10].
Igor Prodan has reported the model which portrays influence of R & D expenditures on number of
patent applications in selected OECD countries and central Europe [4]. Research confirms the
positive correlation between the two. Meo and Usmani have presented impact of R&D spending
on research publications, patents and high technology exports among 47 European countries [3].
This research collected the information regarding per capita GDP, R&D expenditure with the
conclusion that R&D expenditure and research publications are the most significant contributing
factors towards a knowledge economy.
Janodia has compared expenditure on R & D and patents of India among SAARC and BRICS
countries [5]. The study revealed that it is essential to raise R&D expenditure to motivate research
activities leading to innovation, increasing patenting and larger number of publications. Dietmar
reported the effect of R&D spending on its productivity in German manufacturing firms [7]. The
result of the study suggested that spillovers affect industries in a heterogeneous manner. Yet

2
another paper by Kamath and Kamat reported clustering of countries based on their R&D
expenditure and its productivity [6]. Here analysis has done based on the number of research
publications, patent applications and trademarks registered with reference to percentage of GDP
spending on R&D.
In the present communication we report a systematic methodology with R&D datasets as input to
the k-means algorithm to cluster the performers as well as the gloomy ones in the interest of the
public policy at large. The impact of research recourses quantified in terms of R & D expenditure
and the corresponding research output, as measured in terms of the quantity of journal articles
have been taken as two independent data sets for the correlation purpose. In order to visualize the
competitive picture we correlate R&D expenditure and journal publications on a global scale
taking into consideration the individual country performance. We use R data mining tool for more
significant analysis resulting into the clusters having varied performance.
2. DATASETS FOR PRESENT ANALYSIS
In this comprehensive study, two independent sources of data, Research and development
expenditure i.e. % of GDP and Scientific and technical journal articles, are brought together for
the assessment purpose. These datasets are available at Knoema (http://knoema.com ) a free to
use web based public and open data platform launched for the purpose of statistical analysis. This
web based platform uses international open data platforms such as World Bank, IMF, UN and
many more as sources to fetch the data. Expenditures for R&D available on this platform are
current, up to date and include capital expenditures for both public and private sector on creative
work undertaken systematically to increase knowledge and the use of knowledge for new
applications. Knoema implies R&D that includes basic research, applied research, and
experimental development. Scientific and technical journal articles refer to the number of
scientific and engineering articles published in the following fields: physics, biology, chemistry,
mathematics, clinical medicine, biomedical research, engineering and technology, and earth and
space sciences. Snapshot of the dataset used in this study shown in figure 1.With such a
voluminous availability of the data sets it would be really worthwhile to deice the trends and
meaningful performance measures out of the same. Following section presents the methodology
for devising the trends from the data sets.
Figure 1. Snapshot of dataset

3
3. ANALYSIS THROUGH CORRELATION AND CLUSTERING
In order to visualize meaningful performance index relating the R & D budget with the
publication output we adopt the K-means clustering algorithm [8]. The data were analyzed by
using MS-Excel and R data mining tool. In this study 76 countries’ R&D expenditure and
research publication for the years 2011, 2010, 2009, 2008, 2005 and 2000 were included. The
correlation coefficient was calculated to find the imminence of relation between these two
parameters. Correlation between R&D expenditure and quantity of research publications data
demonstrated in table 1. All these 76 countries’ correlation coefficients are shown as Radar chart
in figure 2.
Table 1. Correlation of R&D expenditure and Research publications
Figure 2. Radar chart representing correlation coefficients
This correlated data was further used for designing prominent clusters of countries by applying
unsupervised learning. We undertake the same to uncover the relationship between R&D
expenditure and research publications. R open source data mining tool has been used for
clustering of countries based on their correlation coefficients [9]. The K-means algorithm is used
here for grouping countries according to their similarity. Corresponding output of clustering using
‘R’ is shown in figure 3. Three prominent clusters have been devised implying positive, negative

4
or neutral correlation. K-means clustering resulted into three clusters of size 17, 33 and 21 given
in table 2.
Figure 3. Snapshot of Result of clustering in R
Table 2: Clusters of countries based on correlation data
4. INTRA CLUSTER ANALYSIS AND PERFORMANCE MEASURES
The countries belong to cluster 1 have non - linear relationship between R&D expenditure and
Publication as their correlation coefficient is closer to -1. Negative correlation implies a
completely inverse relationship between the R&D expenditure with number of publications. For
instance Bulgaria and Japan belonging to this cluster with correlation coefficient -0.74 and -0.61
indicate gloomy performance in terms of publication output despite being the heavy investment in
R & D shown in table 3.
Positive correlation i.e. correlation coefficient closer to +1 found in countries belonging to
cluster2 reveals a linear relationship between R&D expenditure and Publication. Intra cluster
analysis shows that as the R&D expenditure increases number of publications also increases in
case of these countries. India, Argentina and Malaysia are the main countries belonging to this
cluster with correlation coefficient 0.82, 0.91 and 0.95. Sample data of these countries belonging

5
to cluster 2 can be further analyzed so as to perceive the closeness of the publications versus the R
& D spending given in table 4.
The weak correlation between R&D expenditure and number of publications is exhibited by
countries belonging to cluster 3. The correlation coefficient is closer to 0 implies almost no
relationship between R&D expenditure and number of publications. For example Kuwait and
Serbia belong to this cluster with correlation coefficient 0.0 and 0.28 shown in table 5.
Table 3. Sample data of countries belonging to cluster
Table 4. Sample data of countries belonging to cluster 2
Table 5. Sample data of countries belonging to cluster 3
5. CONCLUSIONS
Our investigation using the K-means clustering algorithm using R-mining is in fact one of its
kinds systematic modus operandi to device the correlation between two independent data sets for
perceiving the performance metrics for the benefit of the policy makers, scientific community and
the society at large. Since the budget for the R & D is drawing from the tax payers, society
ultimately becomes the major stakeholder of the research. Visualization and analysis technique
such as the reported ones not only correlates independent data sets but also throws light on the
preeminent trends and patterns that the stakeholders ought to know. Our investigation reveals that
for 33 countries out of 76, there exists positive linear correlation between R & D Expenditure and
number of research publications. The impact of research funding structure on quantity of journal
articles is fundamental to an understanding of the functioning of research systems. This question
is not of just aesthetic interest to scientific community. In times of limited funding there are
pressures to implement policies that will affect the distribution of funds and the reported

6
framework will definitely ensure judicious distribution from the apex funding bodies like World
Bank.
REFERENCES
[1] Jacob, Brian A and Lefgren (2011), “Lars, The impact of research grant funding on scientific
productivity”, Journal of Public Economics, 95, pp1168-1177.
[2] McAllister, Paul R and Wagner, Deborah Ann (1981), “Relationship between R&D expenditures and
publication output for U.S. colleges and universities”, Research in Higher Education, 15, 3-30.
[3] Meo SA and Usmani Adnan Mahmood (2014), “Impact of R&D expenditures on research
publications, patents and high-tech exports among European countries”, European Review for
Medical and Pharmacological Sciences, 18, pp1-9.
[4] Prodan Igor, “Influence of Research and Development Expenditures on Number of Patent
Applications: Selected Case Studies in OECD Countries and Central Europe”, Applied Econometrics
and International Development. AEID., 2005, Vol. 5-4.
[5] Manthan D. Janodia, “Research and development spending and patents: where does India stand
among SAARC and BRICS”, Current Science, 2015, 108.
[6] R.S. Kamath & R.K. Kamat (2016), “K-Means Clustering for Analyzing Productivity in Light of R &
D Spillover”, International Journal of Information Technology, Modeling and Computing (IJITMC),
Vol.4, No. 2, pp55-64.
[7] Dietmar Harhoff, “R&D Spillovers”, Technological Proximity and Productivity Growth – Evidence
from German Panel Data
[8] Jiawei Han, Micheline Kamber and Jian Pei (2012), Data Minining Concepts and Techniques, Third
Edition, Elsevier Inc.
[9] R.S.Kamath, R.K.Kamat (2016), Educational Data Mining with R and Rattle, River Publishers Series
in Information Science and Technology, River Publishers.
[10] Bozeman, B., & Melkers, J. (1993), “Evaluating R & D impacts”. Boston: Kluwer Academic.
[11] Gaillard, J. (2010) “Measuring Research and Development in Developing Countries: Main
Characteristics and Implications for the Frascati Manual”. Science Technology & Society, 15(1),
pp77-111. doi:10.1177/097172180901500104
Authors
R.S. Kamath
Dr. R.S. Kamath is Associate Professor in the Department of Computer Studies, Chhatrapati Shahu Institute
of Business Education and Research, Kolhapur, India. She obtained her Bachelors and Masters in Computer
Science from Mangalore University. She received her Ph.D. in Computer Science specialized in Computer
Based Visualization from Shivaji University and completed the same in 2011. Dr. Kamath has to her credit
20 research papers published in reputed national and international journals and presented 11 papers in
national conferences. She has completed two minor research funded by UGC. She is the author of two
books and edited one seminar proceedings. The chapter entitled “Cost Effective 3D Stereo Visualization for
Creative Learning – Virtual Reality in Education” is accepted for Encyclopedia of Information Science and
Technology 4th Edition by IGI Global Publishing. Her areas of research interests are Artificial Intelligence,
Virtual Reality, Soft Computing and Data Mining. She has immense skill of around twelve years in
teaching and research.

7
R.K. Kamat
Dr. R.K. Kamat holds the position of Professor in the Department of Electronics and heads the Department
of Computer Science of Shivaji University, Kolhapur. He is Director of Internal Quality Assurance (IQAC)
of the Shivaji University, Kolhapur. He obtained his Bachalors, Masters and M.Phil in Electronics from the
Shivaji University, Kolhapur. Dr. Kamat gained his Ph.D. specialized in Smart Sensors from Goa
University, Goa. Professor Kamat has published over 70 plus papers in International journals of repute and
presented equal number of papers at National and International Conferences. He has published 10 books
though reputed publishing house such as Springer UK. He has published scholarly literature on the quality
issues in higher education. Five students have been awarded Ph.D. under his guidance and 11 more are
working for their Doctorate. He has been involved with various initiatives of the apex organization at
international level such as IEEE USA and Engineering Education group of Central Quinsland Australia.
Through these organizations he is playing key role in spreading the scholastic culture by organizing
conferences at various international destinations. Through this network he has visited various countries. Dr.
Kamat is a recipient of the Young Scientist Award under the fast track scheme of the Department of
Science and Technology (DST) of Government of India.
S.M. Pujar
Dr. S. M. Pujar is Deputy Librarian in Indira Gandhi Institute of Development Research, Gen AK Vaidya
Marg, Goregaon East, Mumbai. Dr. Pujar has published over 20 plus papers in reputed national and
international journals and presented papers in conferences. His areas of research interests are Digital
Libraries, Information Science, ICT applications for Libraries and Library Automation.

CORRELATING R & D EXPENDITURE AND SCHOLARLY PUBLICATION OUTPUT USING K-MEANS CLUSTERING

More Related Content

What's hot (7)

Viewers also liked (15)

Similar to CORRELATING R & D EXPENDITURE AND SCHOLARLY PUBLICATION OUTPUT USING K-MEANS CLUSTERING (20)

Recently uploaded (20)

CORRELATING R & D EXPENDITURE AND SCHOLARLY PUBLICATION OUTPUT USING K-MEANS CLUSTERING