You are on page 1of 3

FV index can group works in the network to two classes – works with divergence potential and works in

convergence mode. If works exhibiting divergence mode is treated as class A (high potential works) and
rest as class B, then we have to check whether the above two criteria are satisfied.

For that 10 different networks representing 10 different fields are analysed.

t=2013 and T=2017

As benchmark measures of performance, indegree and eigenvector centrality are used.

Performance at time t for class A → avg. indegree of class A (t) and avg. eigenvector of class A (t)

Performance at time T for class A → avg. indegree of class A (T) and avg. eigenvector of class A (T)

Change in performance of class A → avg. indegree of class A (T) - avg. indegree of class A (t)

(i.e., ∆ avg. indegree of class A)

And avg. eigenvector of class A (T)- avg. eigenvector of class A (t)

(i.e., ∆ avg. eigenvector of class A)

Similarly change in performance of class B can also be found.

Now for a network, if avg. indegree of class A (t) > avg. indegree of class B (t)

And avg. eigenvector of class A (t)> avg. eigenvector of class B (t)

Performance of A at t > Performance of B at t.

∆ avg. indegree of class A > ∆ avg. indegree of class B

And

∆ avg. eigenvector of class A > ∆ avg. eigenvector of class B,

Then Change in Performance of A > Change in Performance of B.

For 10 networks representing different fields these are checked and the results are shown below.

a) Case of Indegree
In.avg.2013 In.avg.2017 ∆.In.avg.
Network Class A Class B Class A Class B Class A Class B
BT, eng 1.7651 0.0842 2.0201 0.2821 0.255 0.1979
DM 4.252 0.2962 7.6101 1.7632 3.3581 1.467
IT, eng 2.3364 0.2263 3.25 0.9875 0.9136 0.7612
KM, eng 2.62 0.2498 3.4096 0.8803 0.7896 0.6305
ML, eng 3.3938 0.2448 5.4029 1.2125 2.0091 0.9677
RE 5.7257 0.720325 10.7269 4.139572 5.0012 3.419246
TM 2.6197 0.290541 3.1743 0.714286 0.5546 0.423745
VR 9.1385 0.866254 12.4213 2.471918 3.2828 1.605664
SG 5.153465 0.423193 18.65347 7.100904 13.5 6.677711

b) Case of Eigenvector

Eig.avg.2013 Eig.avg.2017 ∆.Eig.avg.


Network Class A Class B Class A Class B Class A Class B
BT, eng 0.0497 0.0013 0.0577 0.0036 0.008 0.0023
DM 0.0281 0.0011 0.0369 0.0051 0.0088 0.004
IT, eng 0.0201 0.0013 0.0254 0.0047 0.0053 0.0034
KM, eng 0.0228 0.0011 0.0332 0.0041 0.0104 0.003
MLE, eng 0.0151 0.0007 0.0239 0.0034 0.0088 0.0027
RE 0.023 0.0014 0.0303 0.0064 0.0073 0.005
TM 0.0225 0.0016 0.0284 0.0037 0.0059 0.0021
VR 0.0164 0.001 0.0178 0.0022 0.0014 0.0012
SG 0.06935 0.0026 0.0859 0.019 0.01655 0.0164

For all the 10 networks explored, these conditions were satisfied and hence FV index’s importance as a
classifier and as a measure of relative importance is verified.

You might also like