The information age and digital communication have transformed the way we collect, process, store, and display data. However, when dealing with georeferenced information, the challenge remains of representing spatial data in a manner consistent with its logical and semantic context. Conventional spatial tables describe attributes in isolation, failing to capture their meaning within the context in which they are embedded. This work proposes the use of knowledge graphs to integrate spatial and non-spatial databases, aiming to represent semantic relationships between geographic entities and telecommunications infrastructure across Brazil. Open data from Anatel and IBGE were used to structure the knowledge graph, which allows for the analysis of the distribution of radio base stations in each Brazilian municipality and their geoeconomic impact. The study demonstrates the feasibility of enriching analyses of mobile telecommunications infrastructure (2G, 3G, 4G, and 5G), supporting Anatel's decision-making in public policies aimed at ensuring interoperability.
Mobile telecommunications network data is characterized by heterogeneous data types and complex relationships between the entities represented. However, the way in which data is currently available is limited in relation to these aspects. For this reason, a form of storage is needed that enables more eficient and meaningful organization of this data. Knowledge graphs are emerging as a promising alternative for organizing information on telecommunications infrastructure and services in Brazil.
In this paper, we present a proposal for a knowledge graph structure that represents the topology or architecture of mobile telecommunications networks in Brazil. The proposal allows for data scalability and more accurate analyses of the organization and behavior of radio base stations (ERBs) in diferent regions. The proposal also ofers the following advantages: • Knowledge graphs allow visualization of the network topology, understanding the relationship between its physical and logical components. • With graph-based reasoning, it is possible to identify the root cause of failures or service degradation, streamlining the maintenance and restoration process of the telecommunication network. • They allow the use of performance metrics related to network configurations to recommend antenna installation optimization strategies.
The geospatial database was obtained from open data from the IBGE (Brazilian Institute of Geography
and Statistics), including names of states and municipalities, territorial geometries and other geospatial
information about Brazil [
CEUR Workshop Proceedings
ceur-ws.org
ISSN1613-0073
The rest of the document is organized as follows: Section 2 presents related work in telecommunications
that uses knowledge graphs and ontologies; Section 3 describes the proposal for a knowledge graph for
mobile telecommunications networks in Brazil; Section 4 presents examples of queries made to the graph
and discusses its benefits; finally, Section 5 presents conclusions and suggestions for future work.
2. State of the Art
knowledge graphs began to be used in 2012 by Google [
A notable example of this approach can be seen in Google’s search engine. When you search for the name of a city, it not only displays related hyperlinks, but also presents organized information about its location, cultural aspects, health infrastructure, and geographic data. This data comes from an ontologically structured knowledge graph, which links diferent sources of information based on formally established semantic relationships.
In addition to applications in search engines, knowledge graphs have become fundamental tools in areas such as natural language processing, recommendation systems, semantic relation extraction, knowledge engineering, and intelligent chatbots. Their structure facilitates data organization during the preprocessing and modeling stages, making them especially useful for feeding artificial intelligence algorithms with contextualized and semantically enriched data.
The paper Ontology for IP Telephony Networks [
The TOUCA Project proposes the ToCo (Telecommunication CANvas Ontology) ontology [
Advances in the use of knowledge graphs have also been made in the direction of eficiently compressing
the density of data generated by 6G mobile networks, called pervasive multilevel native AI (PML)[
In the works presented above, it is worth highlighting the relationship between telecommunications networks and knowledge graphs to understand the semantic behavior between the physical and logical elements of a network for the construction of new protocols, architectural elements, topologies, and data refinement. In continental countries like Brazil, Russia, and China, where deployment costs are a barrier to its growth, this view of the network helps to build mobile telecommunications networks eficiently.
Structuring spatial data together with non-spatial data is an inherently complex task, reflecting the wide
diversity of formats, such as satellite images, cartography, vector data, matrix data, demographic data,
textual data, and numerical data. Although tables help with organization, they are limited in their ability
to visualize the semantic relationships between cartographic representation, geographic attributes, and
external data from multiple sources. For this reason, geographic information systems (GIS) arose from the
need to capture, manipulate, analyze, and model geographic data and metadata to provide an integrated and
geospatial context-oriented view [
An additional limitation of the conventional tabular model lies in the dificulty of performing complex data queries, thereby reducing the potential to explore semantic relationships between data elements.
To represent the structure of mobile telecommunications networks in Brazil using a knowledge graph, the ifrst step was to define the types of data to be used. Cartographic data in Shapefile format, numerical data, and text extracted from IBGE sources and georeferenced telecommunications data from Anatel’s MOSAICO system were therefore used. Geographic, cartographic, and geoeconomic data from IBGE on Brazilian territory were used to analyze the structure of radio base stations (ERBs) with ANATEL data, interpreting how MOSAICO system mobile services are impacted throughout Brazilian territory.
Based on the understanding of this data, the corresponding knowledge graph was generated, linking the georeferenced data with the geographic data from the two databases. It expresses the relationships between the data of each ERB in relation to the spectrums it controls, its influence in the regional zone, antenna coverage by region, frequencies available for use, types of mobile services, and how each region manages each mobile telecommunications service, all with the organization of the geographic entities described in the IBGE data. Thus, as a basis for constructing the graph, the defined conceptual schemes of ANATEL and IBGE data show how the ERBs are distributed and their impact on each region.
The geospatial data extracted from the IBGE platform are in Shapefile (.shp) format, which contains the
geometric characteristics and territorial boundaries of each Brazilian municipality. From the cartographic
database, 5,572 municipalities and 27 states were extracted, along with six attributes for each municipality
that are implicit in the map. External geographic information on states and municipalities was obtained
from the Cidades@ portal [
The SPECTRUM MOSAICO platform from Anatel is intended to manage telecommunications services, keep track of broadcasting station records, and show whether each ERB’s infrastructure conforms with Anatel regulations. The open data used on licensed telecommunications stations in Brazil is in tabular .csv format with 2,083,046 rows and 41 columns, which separates the characteristics of each station’s antennas. This study considers mobile service stations (2G, 3G, 4G, and 5G), and through their geographic position (latitude and longitude), it is possible to infer the impact of each ERB in the region where it is installed and how they interact with each other.
Using data from IBGE and Anatel, it is possible to establish spatial relationships with mobile architecture
and its relationship with stations, making it possible to infer which antennas are in a specific area, perform
network coverage or saturated area analysis, plan the network, and detect anomalies. In addition, it
is possible to integrate data from other sources to infer the operational, administrative, or regulatory
characteristics of public services. The principle behind this vision is the importance of semantic enrichment
of data from multiple domains [
A radio base station (ERB) is a fixed installation that houses the equipment responsible for enabling communication between mobile devices and the telephone company, providing signal coverage in a geographical area.
Based on current legislation [
So each antenna has a concrete physical infrastructure and uses transmitter equipment to broadcast signals within the coverage area. Antennas installed in critical locations must be identified in order to follow an installation and operation protocol that is diferent from the others. The Figure 1 presents the conceptual model of the ERB structure with the antennas and the technical characteristics of their operation associated with the equipment used in the transmission infrastructure.
To model the information related to licensed ERB concessions, latitude and longitude coordinates were used as a basis for establishing an indirect relationship between stations. This approach prevents data explosion if the ERBs were directly connected to other stations. With the support of thematic cartography extracted from the IBGE map, it is possible to indirectly associate each station with its structural characteristics and infer its area of influence in the geographical region where the ERB provides signal coverage.
Thus, to represent the distribution of ERBs in cities and organize them administratively in accordance with Anatel, only neighbouring cities in the same state were connected directly, and the state module was connected to the capital city, as illustrated in Figure 2. This connection between cities is useful for solving the travelling salesman problem when Anatel’s regional superintendence needs to perform maintenance and inspections of its internal state mobile infrastructure. The states were connected directly to their neighbours to formalize the connections between states, according to IBGE data.
By way of illustration, Figure 3 shows the relationship between spatial data from ERBs. It helps illustrate
the argument that integrating data from the nearest stations with their geographic position contributes
to the study of improving the architecture of mobile telecommunications networks for the provision of
services in the coverage area, with the geographic and geodetic conditions presented in the IBGE data [
Based on the conceptual models presented, a large graph was constructed representing the MOSAICO system ERB data and IBGE data using Neo4J’s CYPHER language. The Figure 4 shows part of the graph referring to the structure of ERB 64300. This figure shows the main mobile technologies: GSM (2G), WCDMA (3G), LTE (4G), and NR (5G); the antennas associated with them; equipment that operates their respective antennas; and which antennas are installed in a new project without legacy infrastructure (Greenfield). By looking at the graph, it is possible to understand the structure of how the ERB operates, how its antennas are connected, and how the spectrum of one antenna afects that of another, knowing at what power it operates.
Given its size and scope of representation, viewing the graph in its entirety is not very informative. For this reason, navigation through the graph is done through specific queries written in the CYPHER language.
In order to evaluate the cohesion of the graph and its possible everyday applications, several queries were performed, such as: • physical and topological structure of ERB 64300 with the number of antennas and their technologies; • antennas with diferent physical infrastructure, including the identification number of each antenna and its classification according to the type of sharing; • Identification of nearby stations and the operators that run them; • distribution of mobile telecommunications infrastructure by city, state, or region.
The queries performed showed how a knowledge graph can be used to model data and extract structured information, both visually (through graphs) and in JSON and GeoJSON formats. More significantly, the queries highlighted how the semantics of the geographic relationships between ERBs can contribute to the extraction of information about physical and logical connections between infrastructure elements.
The challenge of representing spatial data in a manner consistent with its logical and semantic context remains relevant, especially when dealing with georeferenced information. Traditional approaches based on spatial tables often fail to capture the semantic relationships between attributes, limiting the integrated analysis of geographic entities and their interactions.
This work proposed an alternative based on knowledge graphs to integrate spatial and non-spatial databases. Using open data from Anatel and IBGE, it was possible to structure a model that not only represents the distribution of radio base stations in Brazil but also allows for richer analyses of their geoeconomic impact and the interoperability of telecommunications networks (2G, 3G, 4G, and 5G).
The results show that this approach can significantly contribute to public policy decision-making, helping Anatel identify areas requiring infrastructure investments or regulatory adjustments to ensure eficiency and equity in access to telecommunications services.The Future work could explore applying this model in other geographical contexts or sectors, expanding its usefulness for urban and regional planning.
During the preparation of this work, the author(s) used GPT-4, Language Tools and Grammarly in order to: Grammar, spelling and plagiarism check. After using these tool(s)/service(s), the author(s) reviewed and edited the content as needed and take(s) full responsibility for the publication’s content.