We construct reverse
nearest neighbor heat maps [1] for many cities, which can be used for influence
exploration.
In many cases, we need
to determine the location, for example, for an ATM machine, a meetup location
of a group of friends, or for the next pickup location of a taxi driver, so
that the ATM machine can serve as many customers as possible, the meetup
location is convenient for many members, and the taxi driver can potentially
make more money. In such a decision making process, there are many quantitative
and qualitative factors to consider. One important quantitative factor is the
distance to users and existing ATMs or taxis, which we will refer to as
facilities. The distance to users indicates how many users the location will
potentially attract, while the distance to existing facilities indicates the
degree of competition.
An effective method to
model the above quantitative factor is the reverse nearest neighbor of the
location. There are many existing algorithms that can return the locations with
the largest reverse nearest neighbor set. However, this implicitly assumes that
this quantitative factor is the only factor we consider in the above decision
making process, and it also assumes that this quantitative factor is only
measured by the size of the RNN set, which is generally not true in many real
applications. In fact, the quantitative factor could be any realvalued
function on the RNN set. For example, the function can take into account the
demographic information of the users in the RNN set, or consider social ties
between these users, or model similar attributes of the facilities. In a real
decision making process, apart from these quantitative factors, there are also
many qualitative factors that cannot be effectively quantified or are subject
to decision maker's judgments, such as the area safety or convenience of public
transportation.
Therefore, we propose
to build a heat map, like this, which shows the quantitative measure, which we
call the influence, of each location, and a darker color means a higher
influence. Such heat map allows the decision maker to explore the whole area
while considering various qualitative factors. With a systematically built heat
map, we can also support decision makers with many interactive operations like
selecting out locations with influence above a given threshold.
[1] Yu Sun, Rui Zhang, Andy Yuan Xue, Jianzhong Qi, Xiaoyong Du.
Reverse Nearest Neighbor Heat Maps: A Tool for Influence Exploration, Proceedings of the 31th IEEE International
Conference on Data Engineering (ICDE) 2016.




ADELAIDE [138.4648,138.8589] [34.7303, 35.0705] 







ATHENS [23.6285, 23.8081] [37.9076, 38.0457] 







BEIJING [116.0692, 116.6707] [39.6734, 40.1642] 







CANBERRA [148.9465, 149.2767] [35.1693, 35.4433] 







DARWIN [130.8149, 130.8595] [12.4299, 12.4730] 







DETROIT [83.1291, 82.9918] [42.2778, 42.3854] 







HELSINKI [24.8394, 25.1731] [60.1022, 60.2739] 







HOLLYWOOD [80.2558, 80.1085] [25.9398, 26.0697] 







MANDURAH [115.6064, 115.8522] [32.4652, 32.6868] 







MELBOURNE [144.7559, 145.0992] [37.6634, 37.9339] 







ROMA [12.4321, 12.5174] [41.8859, 41.9490] 







SYDNEY [150.9364, 151.3078] [33.7489, 34.0635] 







TEHRAN [51.3388, 51.5152] [35.6082, 35.7473] 







SHENZHEN [113.7971,114.0738] [22.4942,22.7471] 


