Page 85 - Contributed Paper Session (CPS) - Volume 5

CPS1060 Taik Guan T. et al.
The features of the agriculture land size, population size, population
density, labor force, number of secondary school students, average monthly
rainfall and average daily temperature, have low co-relationship impacts
(below 0.200) to GNI per capita. Fixed broadband (0.746), wireless broadband
(0.712) and telephony services (0.665) are key elements in ICT ecosystems that
boost the socioeconomic status. The research results show that these features
have a high impact on GNI per capita, and the results are in line with the World
Bank and ITU reports which indicate that broadband penetration will
accelerate economic GDP growth.
Life expectancy and birth rate are the two features of social nature with the
highest impact on GNI per capita. Life expectancy has a directional impact
whereas birth rate has reversed directional impact to GNI per capita. It is
observed that telephony penetration has the highest-impact (0.881)
directional relationship in response to fixed broadband penetration. GNI per
capita (0.746), GDP per capita (0.738), life expectancy (0.735) and wireless
broadband penetration (0.678) have moderately high co-relationship impacts
to broadband penetration. Birthrate (-0.741) also has a moderately high
impact co-relationship to fixed broadband penetration but in the reverse
direction. The length of tar road, tourism activity, % of the population with
electricity access, GDP, and GNI have some co-relationship impact to
broadband penetration, with the impact ratings ranging between 0.206 and
0.499. Those features with relatively low impacts to fixed broadband
penetration are population density, land size, agriculture land size, labor force,
average monthly rainfall, population size, number of secondary school
students and average daily temperature.
Life expectancy and birth rate are the two features of social nature with the
highest impact on fixed broadband penetration. The birthrate is the only
feature that has a significant impact in the reversed direction. The result of the
coefficient correlations can serve as a guideline when applying the empirical
model to further study the game theory in public-private-partnerships (PPP).
The data for geographical conditions with medium to high impact, if missing,
will affect the distortion of the empirical model. In other words, if the
collaborative efforts are put in to improve the data points of those features
with high impact, together the PPP will also improve the socioeconomic
response in favor of broadband investment. In this research, GA successfully
generated 64,200 samples. Each sample is labelled with a fitness value which
virtually represents the socioeconomic status. The fitness breakpoint is the
fitness value that separates the label for the generalized datasets. Through the
training and cross-validation process, it is observed that the SVM has been
trained with high accuracy, ranges from 96% to 99% depending on the kernel
used. The cross-validation accuracy is better when the machine is trained with
a larger sample size.

74 | I S I W S C 2 0 1 9