Abstract:
Studying and mapping population mobility pattern in the city dynamically plays a vital
role and has practical applications to urban planning and to plan better transport system.
For long times the city authorities used traditional census and survey data to plan the city
and allocate basic public facilities. Census and survey data are potentially weak to
delineate the dynamic nature of the city as the data are static and generated within long
time interval and decades. Former local researches in population distribution across the
city were based on traditional data sources. Therefore to overcome the pitfalls of
traditional data such as census and survey data we used the seven consecutive days Call
detail record (CDR) data to analyze and map population mobility pattern. The massive
CDR data that accounts 6,702,111 records was collected from Ethio telecom main data
center. The data contained 564,721 and 543,691 distinct subscribers and devices. We
used Jupyter notebook tool of (anaconda 3) environment of Python3.11 version. Arc map
3.10 had been used to develop a model. In order to identify the geographical location of
cellular network tower and to reveal the call activity intensity of a certain area in a timely
and spatial manner we used voronoi location algorithm. We also used K-mean clustering
algorithm comparatively to justify the result found using voronoi algorithm. We carried
out two major experiments in our work .In the first experiment we have used voronoi
location algorithm technique. In the second experiment, we conducted K-mean
clustering technique on the data set. Finally, we got satisfactory result as the two
experiments yield the same result. In both experiments we have systematically identified
cellular network towers in the study area that hosted high call activity intensity and low
call activity intensity on both weekdays and weekends. It in turn enabled us to revealed
the ambient (daytime) population concentration of areas of the city across time and space.
Keywords: Population mobility, Pattern, Call detail records, K-mean clustering