Estimate communication service demand

Abstract

This document discloses a solution for estimating network service capacity requirements in a region of interest. According to one aspect, a computer-implemented method includes: forming a social media layer using one or more social media applications, the social media layer storing records of multiple locations in the region of interest; forming a geographic location layer mapping locations to geographic locations using at least one source storing the real geographic locations of the locations; classifying the locations into multiple categories and assigning weights to each location indicating service capacity requirements dependent on the location category; constructing a capacity layer for the region of interest based on the real geographic locations of the locations provided by the geographic location layer and the service capacity requirements of each location indicated by the weights, the capacity layer indicating the spatial distribution of network service capacity requirements in multiple sub-regions of the region of interest, the multiple sub-regions including sub-regions having locations between them and at least one of the locations.

Description

Technical Field

[0001] The various embodiments described herein relate to the field of wireless communications, and more specifically to estimating capacity requirements or traffic density in a region of interest. Background Technology

[0002] Network planning, used to construct or modify cellular network infrastructure, aims to estimate capacity requirements in a region. Population density in that region can be used as input in network planning, but this density is not necessarily related to capacity requirements. For example, this density may be unevenly distributed among customers of different network operators. During network operation, network planning can be carried out by monitoring certain key performance indicators (KPIs) that indicate capacity and reacting to the results of these KPIs. Summary of the Invention

[0003] Some aspects of this invention are defined by the independent claims.

[0004] Some embodiments of the present invention are defined in the dependent claims.

[0005] Embodiments and features described in this specification that are not within the scope of the independent claims (if any) are to be interpreted as examples useful for understanding various embodiments of the invention. Some aspects of this disclosure are defined by the independent claims.

[0006] According to one aspect, an apparatus is provided, comprising: at least one processor; and at least one memory including computer program code, wherein the at least one memory and the computer program code are configured, together with the at least one processor, to cause the apparatus to perform the following operations: forming a social media layer by using one or more social media applications, the social media layer storing records of a plurality of locations in a region of interest; forming a geolocation layer mapping locations to real geolocations by using at least one source storing the real geolocations of the locations; classifying the locations into a plurality of categories and assigning weights to each location indicating service capacity requirements depending on the category of the location; and constructing a capacity layer for the region of interest based on the real geolocations of the locations provided by the geolocation layer and the service capacity requirements of each location indicated by the weights, the capacity layer indicating the spatial distribution of network service capacity requirements in a plurality of sub-regions of the region of interest, the plurality of sub-regions including sub-regions having sub-regions between the locations and at least one of the locations.

[0007] In one embodiment, at least one memory and computer program code are configured, together with at least one processor, to enable the device to output a capacity layer for network planning of a cellular network infrastructure.

[0008] In an embodiment, at least one memory and computer program code are configured, together with at least one processor, to cause the means to: combine social media data representing the same location into the same record in a social media layer, the social media data being obtained from at least a first data source and a second data source, wherein social media data obtained from the first data source is considered to represent the same location as social media data obtained from the second data source if the social media data all indicate locations within a certain distance from each other, and / or if the social media data all indicate the same location by means of names with a similarity higher than a determined threshold.

[0009] In one embodiment, at least one memory and computer program code are configured, together with at least one processor, to enable the device to build the capacity layer without measurement data indicating network traffic volume.

[0010] In one embodiment, at least one memory and computer program code are configured, together with at least one processor, to enable the device to plan network service capacity requirements for sub-areas between at least two locations based on the weights of at least two locations.

[0011] In one embodiment, at least one memory and computer program code are configured, together with at least one processor, to enable the device to construct the capacity layer using the following information as training input: a social media layer storing records of other locations in an additional region of interest, and for each of the additional locations, one of a plurality of categories indicating the business capacity demand of the respective location; and business capacity demand measured at the additional region of interest.

[0012] In one embodiment, the region of interest represents a first city, and the other region of interest represents a second city that is different from the first city but is determined to have social media activities related to the social media activities of the first city.

[0013] In one embodiment, at least one memory and computer program code are configured, together with the at least one processor, to cause the device to construct the capacity layer by at least the following operations: dividing the region of interest into the plurality of sub-regions using a geographic location layer; calculating the service capacity requirements for each sub-region of the plurality of sub-regions by using one or more weights of one or more locations in one or more locations in one or more sub-regions that are neighbors of each sub-region of the plurality of sub-regions.

[0014] In one embodiment, at least one memory and computer program code are configured, together with at least one processor, to enable the device to further assign weights based on the demographic type of a sub-region of the corresponding location.

[0015] In one embodiment, the weight associated with higher business capacity demand is assigned to a specific category in a sub-region with an urban demographic type, but not to a specific category in a sub-region with a suburban or rural demographic type, while the weight associated with lower business capacity demand is assigned to another specific category in a sub-region with an urban demographic type, but not to the other specific category in a sub-region with a suburban or rural demographic type.

[0016] In an embodiment, at least one memory and computer program code are configured, together with the at least one processor, to enable the device to acquire information about at least one of building types and building sizes in the region of interest, and to use the information about at least one of building types and building sizes as other input when constructing the capacity layer.

[0017] In one embodiment, at least one memory and computer program code are configured, together with at least one processor, to enable the device to store addresses of multiple locations obtained from one or more social media applications in a social media layer, map the addresses to geographic coordinates in a geographic location layer, and map the multiple locations to sub-regions using the geographic coordinates.

[0018] According to one aspect, a computer-implemented method is provided for estimating network service capacity requirements in a region of interest, comprising: forming a social media layer by using one or more social media applications, the social media layer storing records of multiple locations in the region of interest; forming a geographic location layer mapping locations to geographic locations by using at least one source storing the real geographic locations of the locations; classifying the locations into multiple categories and assigning weights to each location indicating service capacity requirements depending on the category of the location; constructing a capacity layer for the region of interest based on the real geographic locations of the locations provided by the geographic location layer and the service capacity requirements of each location indicated by the weights, the capacity layer indicating the spatial distribution of network service capacity requirements in multiple sub-regions of the region of interest, the multiple sub-regions including sub-regions having locations between them and at least one of the locations.

[0019] In one embodiment, the computer-implemented method further includes using a capacity layer in the network planning of the cellular network infrastructure and selecting the location of cells for the cellular network infrastructure.

[0020] In one embodiment, the computer-implemented method further includes combining social media data representing the same location into the same record in a social media layer, the social media data data source being obtained from at least a first data source and a second data source, wherein the social media data obtained from the first data source represents the same location as the social media data obtained from the second data source if the social media data all indicate that the location is within a certain distance from each other, and / or if the social media data all indicate the same location by means of names with similarity above a determined threshold.

[0021] In one embodiment, the capacity layer is constructed without measurement data indicating network traffic volume.

[0022] In one embodiment, the network service capacity requirements are planned for a sub-region between the at least two locations based on the weights of the at least two locations.

[0023] In one embodiment, the following information is used as training input to construct the capacity layer: a social media layer storing records of other locations in an additional region of interest, and for each of the other locations, one of a plurality of categories indicating the business capacity demand of the corresponding location; and the business capacity demand measured at the additional region of interest.

[0024] In one embodiment, the region of interest represents a first city, and the additional region of interest represents a second city that is different from the first city but is determined to have social media activities related to the social media activities of the first city.

[0025] In one embodiment, the capacity layer is constructed by at least the following operations: dividing the area of ​​interest into multiple sub-regions using a geographic location layer; calculating the business capacity requirements for each sub-region in the multiple sub-regions by using one or more weights of one or more locations in one or more locations in one or more sub-regions that are neighbors of each sub-region in the multiple sub-regions.

[0026] In one embodiment, the assignment is further based on the demographic type of the sub-region at the corresponding location.

[0027] In one embodiment, the weight associated with higher business capacity demand is assigned to a specific category within a sub-region with an urban demographic type, rather than to a specific category within a sub-region with a suburban or rural demographic type, while the weight associated with lower business capacity demand is assigned to another specific category within a sub-region with an urban demographic type, rather than to the other specific category within a sub-region with a suburban or rural demographic type.

[0028] In one embodiment, information about at least one of the building type and building size in the region of interest is obtained and used as other inputs when constructing the capacity layer.

[0029] In one embodiment, the computer-implemented method further includes storing addresses of multiple locations obtained from one or more social media applications in a social media layer, mapping the addresses to geographic coordinates in a geographic location layer, and mapping the multiple locations to sub-regions using the geographic coordinates.

[0030] According to one aspect, a computer program product is provided, embodied on a computer-readable medium and comprising computer-readable computer program code for a first wireless network, wherein the computer program code configures a computer to perform a computer process for estimating network traffic capacity requirements in an area of ​​interest, comprising: forming a social media layer by using one or more social media applications, the social media layer storing records of a plurality of locations in the area of ​​interest; forming a geographic location layer mapping locations to geographic locations by using at least one source storing the real geographic locations of the locations; classifying the locations into a plurality of categories and assigning a weight to each location indicating traffic capacity requirements depending on the category of the location; constructing a capacity layer for the area of ​​interest based on the real geographic locations of the locations provided by the geographic location layer and the traffic capacity requirements of each location indicated by the weights, the capacity layer indicating the spatial distribution of network traffic capacity requirements in a plurality of sub-regions of the area of ​​interest, the plurality of sub-regions including sub-regions having locations between locations and sub-regions having at least one of locations.

[0031] In one embodiment, the computer program product further includes computer program code for configuring the computer to perform any of the above embodiments of the computer-implemented method. Attached Figure Description

[0032] The following description of embodiments is by way of example only, with reference to the accompanying drawings, wherein...

[0033] Figure 1 The diagram illustrates cellular network planning;

[0034] Figure 2 An embodiment of a process for estimating communication service capacity requirements for a region of interest is illustrated;

[0035] Figure 3 The diagram shows Figure 2 Some embodiments of the process;

[0036] Figure 4 The diagram illustrates how to construct a capacity layer by dividing the region of interest into a grid;

[0037] Figure 5 The diagram illustrates training a machine learning algorithm and using the trained machine learning to build a capacity layer.

[0038] Figure 6 and Figure 7 The diagram illustrates the structure and operation of machine learning; and

[0039] Figure 8 A block diagram illustrating the structure of a device according to an embodiment of the present invention is shown. Detailed Implementation

[0040] The following embodiments are examples. Although the specification may refer to "a," "an," or "some" (or more) embodiments in several places, this does not necessarily mean that each such reference points to the same (or more) embodiments, or that the feature applies only to a single embodiment. Individual features of different embodiments may also be combined to provide other embodiments. Furthermore, the words "comprising" and "including" should be understood not to limit the described embodiments to consisting only of those features mentioned, and such embodiments may also include features / structures not specifically mentioned.

[0041] Figure 1 The illustration depicts the concept of cellular coverage for a cellular communication system. A cellular communication system can be any system, ranging from second-generation systems such as GSM to modern systems such as Advanced Long Term Evolution (LTE-A, LTE-Advanced) or New Radio (NR, 5G), without limiting the embodiment to any particular architecture. The embodiment can also be applied to other types of communication networks. Some examples of other options suitable for the system are Universal Mobile Telecommunications System (UMTS) Radio Access Network (UTRAN or E-UTRAN), Long Term Evolution (LTE, the same as E-UTRA), Wireless Local Area Network (WLAN or WiFi), Global Microwave Access Interoperability (WiMAX), Wideband Code Division Multiple Access (WCDMA), or any combination thereof.

[0042] refer to Figure 1One aspect of cellular network planning is providing coverage in every part of the Area of ​​Interest (AOI), for example, where cellular network operators choose to provide coverage. Cities and other densely populated areas are of particular interest due to the large number of users providing communication services. In these areas, there may be a particular focus on providing coverage that meets the demand for communication services and sufficient communication service capacity. Nowadays, people are mobile, so estimating the service demand of an area is not easy. At the level of macrocells 100, 102, 104 covering large areas, estimation is easier because the larger the area, the more stable the service demand. However, estimating the service demand of smaller areas (such as buildings or each building in a city) is both time-consuming and inaccurate. It is time-consuming in the sense that each building must be evaluated individually. It is also inaccurate in the sense that the mobility of people within small areas is unpredictable. Therefore, estimating the demand of small cells 110, 112, or hotspots within macrocells requires greater precision.

[0043] The geographical coverage area of ​​a cellular communication system can include multiple different types of radio cells. Radio cells can include macrocells 100, 102, 104 (or umbrella cells), which are large cells, typically with diameters of up to tens of kilometers. Within macrocells, smaller cells, such as microcells, femtocells, or picocells, can be built to provide local hotspots to improve service capacity. Modern cellular communication systems can be implemented as multi-layered networks comprising various cell types. Typically, in a multi-layered network, one access node or base station (eNB, gNB in ​​some system specifications) provides one or more cells of a certain type, and multiple access nodes or base stations are required to provide such a network structure.

[0044] Figure 2 An embodiment of a computer implementation of a process for estimating communication service requirements in a region of interest is illustrated. (Reference) Figure 2The process includes the following steps, performed by at least one processor or computer processing system: forming a social media layer (block 200) using one or more social media applications, which stores records of multiple locations within the region of interest; forming a geographic location layer (block 202) mapping locations to geographic locations using at least one source storing the real geographic locations of locations; classifying locations (block 204) into multiple categories and assigning a weight to each location, the weight indicating the service capacity requirements depending on the category of said location; and constructing a capacity layer (block 204) for the region of interest based on the real geographic locations of the locations provided by the geographic location layer and the service capacity requirements of each location indicated by the weights, the capacity layer indicating the spatial distribution of network service capacity requirements in multiple sub-regions of the region of interest, including sub-regions with locations between locations and sub-regions with at least one location. Various social media applications exist that allow mapping a user's location to various locations such as cafes, restaurants, hotels, etc. Examples of such applications include Facebook, Twitter, Foursquare, Google Places, Yellow Pages, etc. Such applications can be used as sources of social media data, allowing estimation of the location of people and the distribution of people. They typically provide mappings between people and locations, as well as location indications of locations. However, depending on the application, the location may have low geographic accuracy. Some applications use satellite positioning to provide the site's location. Satellite positioning can be inaccurate, depending on the user's location and the system used to perform the positioning. Positioning errors can be tens of meters or even more, making these sources unsuitable for estimating business needs in very small AOIs, such as 100 square meters. Some social media applications allow the registration of the location's address, but the registered address may be inaccurate in terms of its true geographic location, for example, due to the inaccurate geolocation capabilities of the social media application. Figure 2 The implementation uses a geographic layer to bind locations to the correct geographic locations (coordinates) by using a more accurate data source of the location's location. For example, such a source could indicate the address of each location on a geographic map that can be mapped to an AOI. Another data source for the geographic layer could be a navigation application that stores the location's location. An example of such a navigation application is... Applications. Another example of a data source for the geographic layer is Google Places, which provides more accurate geographic locations than Facebook, Twitter, Foursquare, and others. By binding locations to their correct geographic locations, business capacity requirements can be estimated more accurately, thus providing a more accurate capacity layer.

[0045] In this embodiment, if the exact location of a place is determined from a social media data source, block 202 can utilize the location obtained from the social media data source. An example of such an accurate source is Google Places.

[0046] In this embodiment, the capacity layer is used for network planning of cellular network infrastructure. In this regard, the capacity layer can be understood as representing the average or static distribution of service capacity demand in an area of ​​interest. This can be understood as a constant, periodic, or other regular trend in service capacity demand, thereby facilitating network planning. The location of cells, particularly small cells, can be determined based on the capacity layer. However, the capacity layer can also be used for other purposes.

[0047] In this embodiment, the social media layer stores the number of users who have checked in at each location using one or more social media apps. The number of check-ins can be used as input to assign weights to that particular location. Check-ins can be understood through comparisons of social media apps. If someone writes a review of a location on a social media app, such as TripAdvisor, that person has checked in at that location. If that person has indicated a visit to a location on a social media app such as Facebook, Twitter, or Foursquare, that person has checked in at that location. Many social media apps track the number of check-ins for each location that can be directly used as input.

[0048] In this embodiment, a capacity layer is constructed in the absence of measurement data indicating network traffic volume measured in one or more wireless networks within the AOI. The measured traffic can be used as additional input for assigning weights, such as for network planning and determining cell locations, or it can be used for training the execution of Block 204, but this is not mandatory. A social media layer provides indications of people distribution within the AOI, while a geolocation layer binds locations to the correct geolocations. Indications of people distribution may include check-in information about locations within the AOI, ratings of locations within the AOI, opening hours of locations within the AOI, etc.

[0049] As mentioned above, locations can be categorized. Categories can define the economic, public, or social function of a location, such as a coffee shop, restaurant, church, residential building, office, government building, etc. This can be used as additional input when building social media and capacity layers. For example, the Google Places application programming interface (API) is a tool for finding such locations. It also provides a category for locations and can therefore be useful for categorization. This categorization can be used to weight the business capacity demand for each user in a location or area of ​​interest. For example, a certain number of check-ins and / or ratings might set a different business capacity demand for a coffee shop than for a church. As another example, a category can indicate the business capacity demand in the absence of observed check-in numbers. In this case, the weighting might rely on statistical assumptions, measurements, or observations for each category. For example, a certain business capacity demand can be assigned to a coffee shop, and a different business capacity demand can be assigned to a church, etc. Each category can have a unique weight for business capacity demand, and by using the locations and weights of a sufficient number of locations in the AOI, the statistical significance of the total business capacity demand and its distribution can be derived for that AOI.

[0050] The social media layer can be understood as a database or data structure that stores social media data that links users to locations. The data structure may include records for each location, including the location's name or another identifier and the number of users linked to that location. Additionally, the record may store other information obtained from the social media application and related to the location. Such other information may indicate the business capacity requirements at the location's location, such as the business capacity per user, and may include user ratings for the location and / or the location's opening hours. The social media layer can be built upon social media data obtained from publicly available sources, and / or it may be built upon commercially available social media data.

[0051] Once the capacity layer has been built, the social media layer may be removed. Some sources may specify time limits for the social media data they acquire. In any case, the purpose of the social media layer may be to build the capacity layer, and once the capacity layer is built, the social media layer may no longer be needed and can be discarded.

[0052] Then let's refer to Figure 3 To describe Figure 2 Some embodiments of the process. Figure 3 Various embodiments of blocks 200 to 204 are described, and these embodiments are considered independent of each other. References Figure 3Social media data can be stored in various databases, each specifically for each social media app. Examples include Google Places database 300, Facebook database 302, and Foursquare database 304. (Execute...) Figure 3 The computer system used in the process can access these databases and obtain social media data that links users to locations. Such social media data may include check-in data, user ratings for each location, and / or other information indicating that a user has visited the location at that geographic location. As a result of obtaining social media data from the databases, the computer system can build a database that includes records for each discovered site.

[0053] Various social media applications and databases may store the same or similar information under different names or categories. For example, the name for a location in one database may differ from the name for the same site in another database. Furthermore, different databases may use different classifications for the same location. Additionally, different databases may store the location as an address or geographic coordinates, but as mentioned above, the accuracy of the location may vary. Therefore, a database at this stage may include duplicate records for the same location, and these records may store the same or different information about the same location. In block 310, the computer system may combine social media data representing the same location but collected from different social media sources. In one embodiment, the computer system combines social media data representing the same location into the same record in social media layer 305. Fuzzy logic or other schemes may be used in the combination. The combination may follow the following principles: If both social media data indicate that the location is within a certain distance of each other and / or if both social media data indicate the same location through names with a similarity higher than a determined threshold, then the social media data obtained from the first data source can be considered to represent the same location as the social media data obtained from the second data source. The similarity of location names can be required to be, for example, at least 70%, at least 80%, or 90%, and thresholds can be set accordingly. For example, a specific distance can be determined based on the average estimated inaccuracy of the least accurate social media sources. Examples include less than 80 meters, less than 70 meters, or less than 60 meters. Using both conditions (name and distance) naturally improves the performance of combining records that truly represent the same location. Combining can include unifying the name of the location (block 312), for example, retaining the official name of the location while discarding other names. Examples of combination results are illustrated in Table 1 below. Clearly, the number of locations can be significantly higher.

[0054]

[0055] Table 1

[0056] As illustrated in Table 1, social media data can include a variety of information for different locations, and some information present for one location may be missing from records for another location. One embodiment of the process includes block 314, where the computer system predicts the missing information for a location by using available information for that location and / or for other locations as input (e.g., input to a random forest algorithm). Alternatively, another algorithm can be used. This algorithm can fill in the missing information based on information available at other sites. For example, the filling logic can use the average of other locations of the same type and with similar other characteristics. For example, the Google rating of Flanco Cafe can be calculated as the average of the Google ratings of other cafes in the same or similar area as Flanco Cafe. Similar areas can be understood in the context of similar socioeconomic or demographic categories in that area, such as “urban,” “suburbs,” “rural,” etc. As a result of block 314, the social media layer 305 of Table 1 can be modified to the form of Table 2, where at least some of the missing information is filled in using predictions.

[0057]

[0058] Table 2

[0059] After executing block 310, the computer system can store or update the social media layer in database 305, which stores the social media layer.

[0060] As a result of block 202, a database as shown in Table 3 below can be constructed for the AOI. Location coordinates can be obtained from sources that provide accurate location data for locations included in the social media layer, such as HERE or Google Places. In some embodiments, the social media layer can easily include coordinates obtained from social media data sources. However, as mentioned above, such coordinates may be inaccurate. Therefore, block 202 can be executed to correct this inaccuracy. Block 202 can be used to reconcile location coordinates in the social media layer that are easily obtained from multiple social media sources. For example, Google Places can provide accurate location coordinates for a location by using the location's actual geographic location, while Facebook can provide inaccurate location coordinates for the location obtained from a satellite positioning receiver. Block 202 can then discard location coordinates deemed inaccurate and maintain only the accurate location coordinates for the location. Discarding can be based on prior knowledge of the social media application or by categorizing those that provide sufficient accuracy in location coordinates and those that do not. In the case of missing accurate location coordinates, external geolocation sources (such as HERE) can be used to provide accurate location coordinates for (multiple) locations.

[0061]

[0062] Table 3

[0063] Block 204 can then combine Tables 2 and 3, and thus provide a capacity layer, for example, Table 4. Table 4 can be constructed based on the combined Tables 2 and 3 using a machine learning model configured to analyze the information stored in the combined Tables 2 and 3. The machine learning model can be trained using real measurement data, as described in some of the embodiments below.

[0064]

[0065] Table 4

[0066] Table 4 illustrates the service capacity requirements of sub-regions within an AOI, as described below. In the embodiment of Table 4, a three-dimensional capacity layer considering the floors of a building is constructed. Another embodiment does not include the third dimension (floors). Service capacity requirements can describe service weights relative to coordinate locations. Therefore, the capacity layer can indicate the relative service capacity requirements of individual sub-regions within the AOI. The number of people in the AOI, for example, can then be distributed across sub-regions to convert relative capacity requirements into absolute requirements, for example, for use in cellular network planning, such as small cell layout design. The size of the sub-regions can be selected based on the purpose of the capacity layer. For example, if the goal is network planning for small cells, the size of the sub-regions may be smaller (e.g., 20 meters × 20 meters) than if the goal is network planning for macrocells (e.g., 100 meters × 100 meters).

[0067] The Area of ​​Interest (AOI) can be divided into multiple sub-regions, and the service capacity requirements for each sub-region can be calculated, as illustrated in Table 4. For this purpose, block 320 may include retrieving locations located within the AOI from a geolocation database and mapping these locations (block 322) to the sub-regions. For example, the set of sub-regions may form a grid covering the (entire) AOI. Therefore, each location in the AOI is assigned to a sub-region. Each sub-region can thus be understood as a bin for service capacity requirement estimation, containing multiple locations that contribute to the capacity requirement estimation for that sub-region. As illustrated above... Figure 2 As described, there may be sub-regions comprising one or more locations and sub-regions located between these locations. Some sub-regions may not have any mapped locations, but may still have capacity requirements planned for these sub-regions based on capacity requirements determined based on locations mapped to one or more adjacent sub-regions.

[0068] Subsequently, in block 330, the capacity requirement for each sub-region can be calculated using social media data stored in the social media layer linked to each sub-region (e.g., Table 4). In other words, calculating the capacity requirement for a sub-region (also referred to as a bin) may at least consider the social media data stored in records of those locations mapped to the sub-region (if any). In another embodiment, calculating the capacity requirement for a sub-region may also use information stored in the social media layer associated with a determined number of sub-regions adjacent to that sub-region. Figure 4 This embodiment is illustrated.

[0069] refer to Figure 4 When estimating the capacity requirements of sub-region A, the computer system can consider social media data mapped to locations in sub-region A, and further consider social media data mapped to locations in sub-regions (a1 to a8) surrounding sub-region A. Depending on the embodiment, even sub-regions surrounding sub-regions a1 to a8 can be considered. Therefore, averaging can be achieved. For example, in sub-region A, locations mapped to it are not included. Figure 4 This can be useful in the scenario illustrated in the diagram. However, sub-regions a1 to a8 surrounding this sub-region contain several locations, and therefore may also affect the service capacity demand in sub-region A. The following equation provides an example of estimating the service capacity demand T for each sub-region A:

[0070] T A =Y1*(Number of locations in Category 1) A +Y2*(Number of locations in Category 2) A +...+m A +Y1*(Number of locations in Category 1) a1 +Y2*(Number of locations in Category 2) a1 +...+m a1 +...+Y1*(Number of locations in Category 1) ax +Y2*(Number of locations in Category 2) ax +...+m ax

[0071] Variables Y1, Y2, ... represent weights that can be defined by category, as described above. Parameter m a1 to m ax This represents a constant that can be determined, for example, based on the population or population density of a specific sub-region and / or other factors. Other parameters can be substituted into the equation, such as the number of check-ins per location. The equation above is based on a linear model in a simplified example. In other embodiments, a different regression model is used.

[0072] As mentioned above, social media data indicating the weight of each location within an AOI or sub-region indicates the (relative) business capacity demand within that AOI or sub-region. Furthermore, the number of locations within an AOI or sub-region is also proportional to the business capacity demand within that AOI or sub-region. The number of ratings or check-ins is also proportional to the number of people at each location. A location with a high rating and a high rating count, ranking first, may represent a larger number of users compared to a second-ranked location with a lower rating and a lower rating count. Opening hours can provide the ability to estimate the temporal characteristics of business capacity demand. For example, it can be estimated that locations within an AOI have higher business capacity demand during opening hours than outside of opening hours. These parameters combined provide an indication of the relative number of people within an AOI or sub-region.

[0073] As mentioned above, location categorization can also be considered in block 330. For example, the service capacity requirements of each user can be determined based on the location category. The computer system can assign different weights to the service capacity requirements of each user or each location in different categories. For example, the service capacity requirements of a hotel or cafe can be set higher than those of a church. As mentioned above, time characteristics can also be considered.

[0074] In this embodiment, location weights are further assigned based on the demographic type of the corresponding sub-region. Demographic types can include urban, suburban, rural, etc. For example, a bar or church in an urban area can be assigned a different weight than a bar or church in a rural area. A bar in an urban area may be associated with higher business capacity demand compared to a bar in a rural area. Generally, a location of a specific category located in an urban sub-region can be assigned a weight indicating higher business capacity demand than a location of a specific category located in a suburban sub-region. Similarly, a location of a specific category located in a suburban sub-region can be assigned a weight indicating higher business capacity demand than a location of a specific category located in a rural sub-region. However, it may be observed that some categories have the opposite effect. For example, a church in a specific rural area may be associated with higher business capacity demand than a church in an urban area. This could be due to higher visitor numbers for each church in the rural areas(s) of a particular AOI. Therefore, the weights associated with higher business capacity demand can be assigned to a specific category within a sub-region with an urban demographic type, rather than to a specific category within a sub-region with a suburban or rural demographic type, while the weights associated with lower business capacity demand can be assigned to another specific category within a sub-region with an urban demographic type, rather than to the other specific category within a sub-region with a suburban or rural demographic type. In summary, demographic type and location classification can provide independent variables for business capacity demand.

[0075] The demographic type of an AOI or its sub-regions can be determined in various ways. For example, a demographic database can be provided, in which a city is divided into demographic regions, each specified according to a specific demographic type. Thus, a direct mapping of the AOI and / or its sub-regions can be derived from the demographic database. In another embodiment, the demographic type of an AOI or sub-region is determined based on locations within the region. For example, if an AOI or sub-region (along with a defined number of neighboring sub-regions) includes at least a defined number of bars, cafes, or restaurants, it is determined to be urban or suburban. If an AOI or sub-region includes only residential buildings and a few public or commercial locations, it can be determined to be rural. Thus, the demographic type can be detected by observing the number and classification of locations within the AOI or a specific sub-region. In yet another embodiment, the demographic type is given. For example, if the AOI only covers the city center of a city, the AOI and all its sub-regions are defaulted to urban.

[0076] In this embodiment, a building layer is constructed that provides information about buildings in the AOI. The building layer may include attributes related to the building's height or geographic region, and / or other building size attributes. Building information in the AOI can be used as an indication of capacity demand. For example, a large residential building may have a higher capacity demand than a small house. Some buildings may have less information in the social media layer, but are still associated with high or significant capacity demand. Therefore, this embodiment uses such a building layer or building database 307, which stores information about buildings in the AOI. Information about buildings may include building height, building type (residential, public, government, industrial, commercial, etc.), and building size information such as the land area covered by the building. Information about buildings can be obtained from various sources. For example, information about buildings can be obtained from commercial sources (purchased), free sources (such as open street maps), satellite imagery, etc. Machine learning can be used to adapt the information into a useful form for constructing the capacity layer. For example, machine learning can be applied to satellite imagery to distinguish building size and geographic location, and machine learning can be cross-referenced with the social media layer to determine the type of building. As mentioned above, the social media layer contains information about the type of location in the AOI. If a building, distinguished from one or more satellite images by a machine learning algorithm, is determined to include at least a certain number of locations, as indicated by the social media layer, then the building can be identified as a commercial building. If the building is large and contains fewer than a certain number of locations, it can be identified as a residential building unless the social media layer or other information otherwise links the building type. The computer system executing block 330 can then use the building information from the AOI as another input for estimating capacity demand, thereby providing a more accurate estimate of capacity demand.

[0077] When mapping information about buildings to business capacity, the machine learning used in database 305 or execution block 330 can utilize the logic described above. Larger buildings are mapped to higher capacity requirements compared to smaller buildings. Each building type can be mapped to a specific capacity requirement for each user, and the number of users can be estimated using the building's size and the social media layer associated with the building, thus providing the overall capacity requirement for the building. Other logic can be implemented additionally or alternatively.

[0078] In this embodiment, buildings are used to form a "grid" and replace the above-described combination. Figure 4 The resulting grid. In this case, buildings can be understood as sub-regions, and locations are mapped to buildings in block 322. Therefore, the building database 307 provides input to block 320.

[0079] In this embodiment, the building database 307 is included in the geographic location database 306. For example, the building database can provide a third dimension (height) to the geographic location database 306. When using Figure 4 In embodiments where the AOI is divided into sub-regions, or generally when the AOI is divided into sub-regions, buildings can also be mapped to sub-regions in block 322. Information about the buildings can then be used to estimate the capacity requirements for each sub-region. For example, in embodiments that calculate capacity layers, the height or size of buildings in each sub-region can be considered. The average building height or size for each sub-region can be calculated for use in capacity layer calculations.

[0080] As described above, a capacity layer can be constructed even without any measurement data related to network traffic in the measurement AOI. In one embodiment, when training is performed... Figure 2Such measurement data can be used when implementing machine learning algorithms for processes at least block 204 or 330. In such an embodiment, the computer system can use the following information as training input to construct a capacity layer: a social media layer for another AOI; and the business capacity demand measured at said other AOI. If there is a correlation between social media activity between the two AOIs, the social media layer of one AOI can be used to train a machine learning algorithm such as a neural network, and thereafter, the computer system can construct a capacity layer based on the social media layer without measurement data. A geographic location layer can be used to improve location positioning. In one embodiment, the AOI represents a first city and the other AOI represents a second city different from the first city. Therefore, when social media activity is correlated between cities, the capacity layer for said other city is also correlated with the actual business capacity demand. For example, social media activity is often similar in different cities within the same country or in similar cities in different countries. Figure 5 These embodiments are illustrated.

[0081] refer to Figure 5 The machine learning algorithm is trained in block 510. The machine learning algorithm can use a neural network or other regression model. Figure 6 An embodiment of a neural network with one hidden layer (e.g., execution block 330) is illustrated, and Figure 7 An example of a computation node in a neural network is illustrated.

[0082] Deep learning (also known as deep structured learning or hierarchical learning) is part of a broader family of machine learning methods based on layers used in artificial neural networks.

[0083] Artificial Neural Networks (ANNs) contain a set of rules designed to perform tasks such as regression, classification, clustering, and pattern recognition. ANNs achieve these goals through a learning process that illustrates various examples of input data and the expected output. Using this, they learn to identify the correct output for any input in the training data manifold. Learning using labels is called supervised learning, while learning without labels is called unsupervised learning. Deep learning typically requires large amounts of input data. In such cases, supervised learning is used.

[0084] A deep neural network (DNN) is an artificial neural network that includes multiple hidden layers 1302 between an input layer 1300 and an output layer 1314. Training a DNN allows it to find the correct mathematical operations so that it can transform the input into the correct output even when the relationships are highly nonlinear and / or complex.

[0085] Each hidden layer 1302 includes nodes 1304, 1306, 1308, 1310, and 1312 for computation. For example... Figure 7 As shown, each node 1304 combines input data 1300 with a set of coefficients or weights 1400, which amplify or attenuate the input 1300, thereby assigning importance to the input 1300 relative to the task the algorithm is attempting to learn. The input-weight products are added 1402, and the sum is passed through an activation function 1404 to determine whether and to what extent the signal should be further processed through the network 330 to influence the final result, such as a classification action. In this process, the neural network learns to recognize the correlation between certain relevant features and the optimal outcome.

[0086] In classification scenarios, the output of a deep learning neural network can be considered the probability of a specific outcome, such as the probability of successful data grouping and decoding in this case. In this context, the number of layers 1302 can vary proportionally to the amount of input data 1300 used. However, when the number of input data 1300 is high, the accuracy of the result 1314 is more reliable. On the other hand, with fewer layers 1302, computation may take less time, thus reducing latency. However, this largely depends on the specific DNN architecture and / or computational resources.

[0087] The initial weights 1400 of the model can be set in various alternative ways. During the training phase, they are suitable for improving the accuracy of the process by analyzing errors in decision-making. Training the model is essentially a trial-and-error activity. In principle, each node 1304, 1306, 1308, 1310, 1312 of the neural network 330 makes a decision (input * weight), and then compares that decision with the collected data to find the differences. In other words, it determines the error and adjusts the weights 1400 based on that error. Therefore, the training of the model can be considered as a correction feedback loop.

[0088] Typically, a stochastic gradient descent optimization algorithm is used to train a neural network model, and backpropagation is used to compute its gradient. Gradient descent attempts to change the weights 1400 to reduce error in the next evaluation, meaning the optimization algorithm is navigating along the gradient (or slope) of the error. If it provides sufficiently accurate weights 1400, then any other suitable optimization algorithm can also be used. Therefore, the post-trained parameters 332 of the neural network 330 can include the weights 1400.

[0089] In the context of optimization algorithms, the function used to evaluate a candidate solution (i.e., a set of weights) is called the objective function. Typically, for neural networks where the objective is to minimize error, the objective function is often referred to as the cost function or loss function. In weight adjustment, any suitable method can be used as the loss function; some examples are mean squared error (MSE), maximum likelihood (MLE), and cross-entropy.

[0090] As for the activation function 1404 of node 1304, it defines the output 1314 of node 1304 given an input or a set of inputs 1300. Node 1304 computes a weighted sum of the inputs, possibly adding bias, and then makes a "activated" or "deactivated" decision based on a decision threshold as a binary activation or using activation function 1404, which gives a nonlinear decision function. Any suitable activation function 1404 can be used, such as sigmoid, rectified linear unit (ReLU), normalized exponential function (softmax), sotfplus, tanh, etc. In deep learning, activation function 1404 is typically set at the layer level and applied to all neurons in that layer. The output 1314 is then used as the input to the next node, and so on, until the desired solution to the original problem is found.

[0091] As an input layer to the neural network, block 510 can use the combined social media layer 500 and geolocation layer 306 for an AOI (AOI 1). As an output layer of the neural network, block 510 can use a capacity layer formed according to the business capacity demand (502) measured for AOI 1. Block 510 can then perform updates to the neural network nodes. The configuration of the neural network thus trained can then be stored in database 504. Similar training can be performed on other machine learning algorithms. The social media layer with corrected geolocations from geolocation layer 306 and the measured (real) business capacity demand can be used to train machine learning. Machine learning can also be trained to determine appropriate weights for each category. As described above, the social media layer 500, together with the geolocation layer, enables the machine learning algorithm to map locations in AOI 1 to determine the spatial distribution of locations in AOI 1. Furthermore, the measured capacity demand 502 can indicate the spatial distribution of business capacity demand in AOI 1. When the number of locations is statistically significant, machine learning can estimate the impact of each category on business capacity demand based on this information, and thus assign appropriate weights to each category.

[0092] The execution of block 512 can be issued when a task to compute the capacity layer is sent to another AOI (AOI 2). Block 512 may include retrieving the configuration of a neural network from database 502 and using the combined social media layer 506 and the geolocation layer 306 for AOI 2 as the input layer to the neural network configured thus. Block 512 may also include measuring the output layer of the neural network to obtain the capacity layer for AOI 2.

[0093] Assuming that social media activity is similar in AOI 1 and AOI 2, the weights determined for AOI 1 based on measurement 502 used as training input can also be considered accurate for AOI 2. Therefore, by leveraging the spatial distribution of locations and their categories in AOI 2 obtained from the social media layer for the geolocation layer of AOI 2, it becomes possible to determine the spatial distribution of business capacity requirements in AOI 2.

[0094] When there are known differences in social media activity between two AOIs, the aforementioned weights can be modified when training a neural network for another AOI. For example, the social media layer for AOI 1, obtained from Database 500, can be modified based on known differences to provide a higher relevance to the actual social media activity of AOI 2. The known differences can be used in the modification to correct for discrepancies in social media activity between AOI 1 and AOI 2. For example, if it is known that users in AOI 2 use the hotel's cellular communication services more frequently than in AOI 1, higher capacity demand weights can be assigned to users in AOI 2 who are linked to the hotel. Similar modifications can be performed for other location categories if there are known differences in behavior between AOIs.

[0095] Another machine learning algorithm uses regression models for capacity demand estimation, for example, to predict users in an area of ​​interest (AOI). As is known in the art, a regression model can be understood as a function...

[0096] Y = X1*m + X2*n + X3*o + X4*p...

[0097] X1 and X2 are variables for a specific sub-region of interest (A) obtained from the table above. X1 could represent the user rating of the top-ranked user, X2 could represent the number of top-ranked users, X3 could represent the user rating of the second-ranked user, X4 could represent the number of second-ranked users, and so on. Y is the estimated parameter (capacity requirement). Therefore, the regression model discovers the weights (m, n, o, p) of each variable by using machine learning and, for example, the measured capacity requirement obtained from database 502 as training input. When the regression model is built on the basis of training, such as when the weights are discovered, the regression model can be stored for later use in the same AOI (with different social media datasets) or in different AOIs for different social media datasets (different values ​​for X1, X2, X3, X4...).

[0098] Figure 8 The diagram illustrates the process of execution. Figure 2 An embodiment of the structure of a device or system for the process of any of its embodiments, or the above-described functions. In one embodiment, the device includes at least one processor, at least one memory, and computer program code configured to cause the device to perform, together with the at least one processor. Figure 2 The process of or any embodiment thereof.

[0099] refer to Figure 8 The device may include a processing system 10, which includes at least one processor. The processing system may include one or more processors from a single physical computer system, or it may include distributed computing resources across various physical computers. The processing system may utilize cloud computing and / or local processing resources.

[0100] The device may include a communication interface 22 configured to provide the device with the ability to communicate over one or more computer networks. For example, the aforementioned database ( Figure 8 26) can be provided in a remote storage resource accessible through a communication interface. The communication interface can be a network adapter that supports one or more network protocols such as Internet Protocol (IP), Transmission Control Protocol (TCP), etc.

[0101] The device may further include a memory 20 that stores one or more computer program products 24 configuring the operation of the device's processor(s). The memory 20 may be implemented using any suitable data storage technology, such as semiconductor-based memory devices, flash memory, magnetic memory devices and systems, optical memory devices and systems, fixed memory, and removable memory. The memory 20 may also store at least some of the databases 26 described above. The memory 20 may also store a configuration database 28 that stores a machine learning agent 18 configured to build a capacity layer according to the embodiments described above. The configuration database 28 may, for example, store the configuration of a neural network (NN).

[0102] The processing system 10 may include a social media layer builder 12, a geolocation layer builder 14, and a capacity layer builder 16 as submodules. Each of modules 12 through 16 may be defined by a separate computer program code module. The social media layer builder 12 may be configured to execute block 200 according to any of the embodiments described above. In some embodiments, builder 12 executes block 310. The geolocation layer builder 14 may be configured to execute block 202, and in some embodiments, block 320. The capacity layer builder 16 may be configured to execute block 204, and in some embodiments, block 330. The capacity layer builder may employ a machine learning agent 18 to construct the capacity layer. For example, the machine learning agent 18 may implement a neural network. In this case, the machine learning agent may be configured to implement... Figure 5 The process, especially blocks 510 and 512.

[0103] As used in this application, the term "circuit system" refers to one or more of the following: (a) a purely hardware circuit implementation, such as an implementation only in analog and / or digital circuit systems; (b) a combination of circuitry and software and / or firmware, such as (if applicable): (i) a combination of (multiple) processors or processor cores; or (ii) a portion of (multiple) processors / software, including (multiple) digital signal processors, software, and at least one memory, which work together to enable a device to perform a specific function; (c) circuitry that requires software or firmware for operation, such as (multiple) microprocessors or a portion of (multiple) microprocessors, even if the software or firmware is not actually present.

[0104] This definition of "circuit system" applies to the use of the term in this application. As a further example, as used in this application, the term "circuit system" will also cover an implementation of only a processor (or processors) or a portion of a processor, such as a core of a multi-core processor and its (or their) accompanying software and / or firmware. For example, and if applicable, the term "circuit system" will also cover specific elements, baseband integrated circuits, application-specific integrated circuits (ASICs), and / or field-programmable grid array (FPGA) circuits used in the device according to embodiments of the invention. Figure 2 , Figure 3 and Figure 5 The processes or methods described in any of its embodiments may also be performed as one or more computer processes defined by one or more computer programs. Individual computer programs may be provided in one or more means of performing the functions of the processes described in conjunction with the accompanying drawings. The computer programs(s) may be in source code form, object code form, or some intermediate form, and may be stored in some kind of carrier, which may be any entity or device capable of carrying the program. Such carriers include transient and / or non-transient computer media, such as recording media, computer memory, read-only memory, electrical carrier signals, telecommunication signals, and software distribution packages. Depending on the required processing power, the computer program may be executed in a single electronic digital processing unit, or it may be distributed across multiple processing units.

[0105] The embodiments described herein are applicable to the computer systems defined above, but are also applicable to other systems. It will be apparent to those skilled in the art that the concepts of the invention can be implemented in various ways as technology advances. The embodiments are not limited to the examples described above, but may vary within the scope of the claims.

Claims

1. A device for communication, comprising: At least one processor; as well as At least one memory, including computer program code, wherein the at least one memory and the computer program code are configured, together with the at least one processor, to cause the device to perform the following operations: A social media layer is formed by using one or more social media applications, which stores records of multiple locations within a region of interest; By using at least one source that stores the real geographic location of the location, a geographic location layer is formed that maps the location to the real geographic location; The locations are classified into multiple categories, and each location is assigned a weight that indicates the business capacity requirements of the category of the location. Based on the actual geographic location of the location provided by the geographic location layer and the service capacity demand of each location indicated by the weights, a capacity layer is constructed for the region of interest. The capacity layer indicates the spatial distribution of network service capacity demand in multiple sub-regions of the region of interest, the multiple sub-regions including sub-regions having locations between them and at least one of the locations.

2. The apparatus of claim 1, wherein the at least one memory and the computer program code are configured, together with the at least one processor, to cause the apparatus to output the capacity layer for network planning of a cellular network infrastructure.

3. The apparatus of claim 1, wherein the at least one memory and the computer program code are configured, together with the at least one processor, to cause the apparatus to: combine social media data representing the same location into the same record in the social media layer, the social media data being obtained from at least a first data source and a second data source, wherein social media data obtained from the first data source is considered to represent the same location as social media data obtained from the second data source if the social media data all indicate that the location is within a certain distance from each other, and / or if the social media data all indicate the same location by means of names with a similarity higher than a determined threshold.

4. The apparatus of claim 1, wherein the at least one memory and the computer program code are configured, together with the at least one processor, to enable the apparatus to construct the capacity layer without measurement data indicating network traffic volume.

5. The apparatus of claim 1, wherein the at least one memory and the computer program code are configured, together with the at least one processor, to enable the apparatus to plan the network service capacity requirements for a sub-region between the at least two locations based on the weights of at least two locations among the locations.

6. The apparatus of claim 1, wherein the at least one memory and the computer program code are configured, together with the at least one processor, to cause the apparatus to construct the capacity layer using the following information as training input: a social media layer storing records of other locations in an additional region of interest, and for each of the other locations, a category of the plurality of categories indicating the business capacity demand of the corresponding location; and business capacity demand measured at the additional region of interest.

7. The apparatus according to any one of claims 1 to 6, wherein the at least one memory and the computer program code are configured, together with the at least one processor, to cause the apparatus to: further assign the weights based on the demographic type of the sub-region of the corresponding location.

8. A computer-implemented method for estimating network service capacity requirements in a region of interest, the method comprising: A social media layer is formed by using one or more social media applications, and the social media layer stores records of multiple locations within the region of interest; A geographic location layer is formed by using at least one source that stores the actual geographic location of the location; The locations are classified into multiple categories, and each location is assigned a weight that indicates the business capacity requirements of the category of the location. Based on the actual geographic location of the location provided by the geographic location layer and the service capacity demand of each location indicated by the weights, a capacity layer is constructed for the region of interest. The capacity layer indicates the spatial distribution of network service capacity demand in multiple sub-regions of the region of interest, the multiple sub-regions including sub-regions having locations between them and at least one of the locations.

9. The computer-implemented method of claim 8 further includes using the capacity layer in network planning of the cellular network infrastructure and selecting the location of cells of the cellular network infrastructure.

10. The computer-implemented method of claim 8, further comprising combining social media data representing the same location into the same record in a social media layer, the social media data data source being obtained from at least a first data source and a second data source, wherein if the social media data all indicate that the location is within a certain distance from each other, and / or if the social media data all indicate the same location by means of names with similarity above a determined threshold, then the social media data obtained from the first data source and the social media data obtained from the second data source represent the same location.

11. The computer-implemented method according to claim 8, wherein the capacity layer is constructed without measurement data indicating network traffic volume.

12. The computer-implemented method according to claim 8, wherein the network service capacity requirements are planned for a sub-region between the at least two of the locations based on the weights of at least two of the locations.

13. The computer-implemented method according to claim 8, wherein the following information is used as training input for constructing the capacity layer: a social media layer storing records of other locations in an additional region of interest, and for each of the other locations, a category of the plurality of categories indicating the business capacity demand of the corresponding location; and business capacity demand measured at the additional region of interest.

14. The computer-implemented method according to any one of claims 8 to 13, wherein the weights are further assigned based on the demographic type of the sub-region of the corresponding location.

15. A computer program product embodied on a computer-readable medium and comprising computer-readable computer program code for a first wireless network, wherein the computer program code configures the computer to perform a computer process for estimating network traffic capacity requirements in a region of interest, comprising: A social media layer is formed by using one or more social media applications, and the social media layer stores records of multiple locations within the region of interest; A geographic location layer is formed by using at least one source that stores the actual geographic location of the location; The locations are classified into multiple categories, and each location is assigned a weight that indicates the business capacity requirements of the category of the location. Based on the actual geographic location of the location provided by the geographic location layer and the service capacity demand of each location indicated by the weights, a capacity layer is constructed for the region of interest. The capacity layer indicates the spatial distribution of network service capacity demand in multiple sub-regions of the region of interest, the multiple sub-regions including sub-regions having sub-regions between the locations and at least one of the locations.

Images