geopolygon: retrieve and correctly reshape the polygon of a geographical area

Summary

Inferring the location of messages, people and events is an increasing branch of researches nowadays. The wide amount of textual data from blogs, microblogs and social media opened to new challenges and, for many application, the location information is needed (Holsapple et al., 2018). However, obtaining the geographic information directly from the messages or users in study is not easy and in many situation it is needed to estimate it. Moreover, especially for social media account information, the geographic indication is often not available at fine-grain resolution but only at Country or State level (Zola et al., 2019a). Thus, to estimate a finer location for a given user/text, some computations and data trasformation are needed.

geopolygon is a Python library able to represent at coordinates level, the original geographic information available at city or regional level. The representation is performed by deriving the polygon area over the World surface to represent the City/Region bounaries. The library have been used in our recent paper (Zola et al., 2019b) for Twitter users geolocation to create a synthetic dataset starting from a frequency distribution of cities over the Globe surface.

The geopolygon Package

The geopolygon Python package provides two services:

1. it retrieves a set of points enclosed in a geographical area; and
2. given the set of points, it computes the polygon.

geopolygon retrieves the set of points from OpenStreetMap (OPS). In order to obtain the data, geopolygon needs two information: the OPS ID and the OPS element. The OPS ID is an unique code able to identify each OPS element.

The OPS element is the basic components of OpenStreetMap's conceptual data model of the physical world. It consists of:

1. nodes: a node is one of the core elements in the OpenStreetMap data model. It consists of a single point in space defined by its latitude, longitude and node id;
2. way: it is an ordered list of nodes;
3. relations: ordered list of one or more nodes, ways and/or relations itself.

For example, a city is a relation composed by ways and nodes. A Country is a collection of relations and it is defined as a relation too.

Thus, given the name of a geographical area, for example the name of a city; geopolygon uses the geopy Python package (geopy) to get the OPS element and the OPS ID. Having the OPS element and the OPS ID, geopolygon retrieves the coordinates of the polygon for the queried area.

geopolygon has been created in order to obtain the polygons of relations (cities, states and countries). Whenever a node or a way is given as input, geopolygon computes an approximated circumference of the area. In order to compute the approxmated circumference geopolygon uses the wikidata tag of the input area to retrive its surface A expressed in . Then, it estimates the radius r as , obtaining the approximated circumference.

In general, the points got from OPS do not have the correct building sequence as shown in Figure 1 (a) about New York city polygon. To avoid this limitation and compute the correct polygon that encompasses the set of points, we applied the concave hull method proposed by Moreira et al. 2007. The algorithm proposed by Moreira et al. is based on a k-nearest neighbours algorithm and it generates the correct concave hull such as exemplified in Figure 1 (b) for the city of New York.

Figure 1:

a: OPS raw polygon	b: Reshaped polygon

Example Usage

The installation of geopolygon is available via Python Package Index (pip), by running the following command:

pip install geopolygon

The main function of geopolygon is area_poly. An example usage of area_poly is reported in the following code:

import geopolygon as gp
city = "Rome"
city_dict = gp.area_poly(city, concave_hull_reshape = 'yes')

print(city_dict.keys())
Out[1]: dict_keys(['type_polygon', 'raw_data', 'processed_data', 'center', 'location_info'])

The output city_dict is a dictionary which contains:

1. type_polygon: can be "Concave Hull" or "Approximate Circumference";
2. raw_data: contains the raw polygon as scrapered from OPS;
3. processed_data: contains the polygon after being processed with the k-nearest neighbours algorithm;
4. center: contains the center of the polygon;
5. location_info: a geopy object which contains all the information about the location.

References

Holsapple, C. W., Hsiao, S. H., & Pakath, R. (2018). Business social media analytics: Characterization and conceptual framework. Decision Support Systems, 110, 32-45.

Moreira, A., & Santos, M. Y. (2007). Concave hull: A k-nearest neighbours approach for the computation of the region occupied by a set of points.

Zola, P., Cortez, P., & Carpita, M. (2019a). Twitter user geolocation using web country noun searches. Decision Support Systems, 120, 50-59.

Zola, P., & Ragno, C., & Cortez, P. (2019b, October). Inferring Twitter users home location based on trend topics. In ASA CONFERENCE 2019 Statistics for Health and Well-being BOOK OF SHORT PAPERS.