Popularity estimation system and geographic feature generation model

Abstract

A popularity evaluation device, which is for evaluating the popularity of specific spots, is provided with: an acquisition unit for acquiring a first map image that includes a first specific spot subject to popularity evaluation and then acquires a plurality of second map images that include a second specific spot, of which a popularity score is already known; a generation unit for generating a first feature vector and a second feature vector by inputting the first map image and the second map images to a geographical feature generation model which receives map images as input and which sends out feature vectors representing geographical features of the inputted map images as output; a score calculation unit that, on the basis of a similarity degree between the first feature vector and the second feature vector, calculates a popularity score for the first specific spot from the popularity score of the second specific spot; and an output unit that outputs the calculated popularity score of the first specific spot.

Description

[Technical field]

[0001] The present invention relates to a popularity estimation system and a geographic feature generation model. [Background technology]

[0002] In a search for facility information that indicates a specific point and is called a POI (Point of Interest), for example, facility information having a facility name that matches an input keyword is output and displayed ranked from the top score. For example, in order to appropriately search for a complex facility that includes multiple facilities and different facilities with the same name, for example, a popularity score that appropriately reflects the popularity of each facility is used to rank the search results. The popularity score is generated based on, for example, the number of times a facility is mentioned in a social networking service (SNS). In addition, regarding the search for facility information, a technology is known that acquires browsing history of restaurant websites and a user's movement history based on GPS, and generates a list of recommended restaurants for a user based on a preference score calculated based on the degree of deviation between them (for example, see Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2013-93015 A Summary of the Invention [Problem to be solved by the invention]

[0004] When generating a popularity score based on the number of times mentioned on SNS etc., the score is calculated based on the history of SNS etc. up to the present. However, there are facilities that should be given a high score even if they are not mentioned on SNS etc. For example, a new facility has no or only a small history of being mentioned on SNS etc., so an appropriate popularity score is not given to the new facility.

[0005] Therefore, the present invention has been made in consideration of the above problems, and has an object to estimate and assign an appropriate popularity score to a POI to which a popularity score has not yet been assigned. [Means for solving the problem]

[0006] In order to solve the above problem, a popularity estimation system according to one embodiment of the present invention is a popularity estimation system that estimates the popularity of a geographically identified specific point, and includes an acquisition unit that acquires a first map image including a first specific point that is the specific point to be estimated, and acquires a plurality of second map images including a second specific point that is a specific point whose popularity score indicating its popularity is known, a generation unit that inputs the first map image and the second map image into a trained geographical feature generation model using machine learning that takes a map image as input and outputs a feature vector that represents the geographical features of the map image, and generates a first feature vector and a second feature vector that represent the geographical features of the first map image and the second map image, respectively, a score calculation unit that calculates a popularity score of the first specific point from the popularity score of the second specific point based on the degree of similarity between the first feature vector and the second feature vector, and an output unit that outputs the calculated popularity score of the first specific point.

[0007] According to the above embodiment, a first feature vector is obtained by inputting a first map image including a first specific point into a geographical feature generation model. The first feature vector represents the geographical features of the first specific point. A second feature vector is obtained by inputting a second map image including a second specific point, the popularity score of which is known, into the geographical feature generation model. The second feature vector represents the geographical features of the second specific point. Then, based on the degree of similarity between the first feature vector and the second feature vector, the popularity score of the first specific point is calculated from the popularity score of the known second specific point, so that it is possible to obtain a popularity score of the first specific point that appropriately reflects the geographical similarity with the second specific point. Effect of the Invention

[0008] It becomes possible to assign an appropriate popularity score to a POI to which no popularity score has been assigned. [Brief description of the drawings]

[0009] [Figure 1] 1 is a block diagram showing a functional configuration of a popularity estimation system including a popularity estimation device according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a hardware block diagram of the popularity estimation device. [Diagram 3] Fig. 3(a) is a diagram showing an example of POI information including attribute information of a POI whose popularity score is known, and Fig. 3(b) is a diagram showing an example of POI information of a POI whose popularity score is unknown and whose popularity score is to be estimated. [Figure 4] FIG. 10 is a diagram illustrating an example of generating a feature vector indicating the geographical features of a POI. [Diagram 5] FIG. 2 is a diagram illustrating a functional configuration of a model generation unit. [Figure 6] FIG. 11 is a diagram for explaining a learning process of a geographical feature generation model. [Figure 7] FIG. 13 is a diagram illustrating a learning process of a popularity estimation model. [Figure 8] Fig. 8(a) is a diagram showing an example of POI information including attribute information of a POI that belongs to a specific set and has a known popularity score, and Fig. 8(b) is a diagram showing an example of POI information of a POI that belongs to a specific set and has an unknown popularity score and is a target for popularity score estimation. [Figure 9] FIG. 13 is a diagram illustrating an example of calculating a second popularity score based on a first popularity score for a POI that belongs to a particular set and has an unknown popularity score. [Figure 10] 13 is a flowchart showing the process of a popularity estimation method in the popularity estimation device. [Figure 11] FIG. 13 is a diagram showing a configuration of a popularity estimation program. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] The following describes an embodiment of a popularity estimation device according to the present invention with reference to the drawings. Whenever possible, the same parts are given the same reference numerals and duplicated explanations are omitted.

[0011] 1 is a diagram showing the functional configuration of a popularity estimation system including a popularity estimation device according to the present embodiment. As shown in Fig. 1, the popularity estimation system 1 includes a popularity estimation device 10, a POI information storage unit 20, a map image storage unit 30, a geographical feature generation model storage unit 40, and a popularity estimation model storage unit 50.

[0012] The popularity estimation device 10 is a device that estimates the popularity of a specific geographically specified point. The specific point is, for example, facility information called a POI. The popularity score indicating the popularity is used in searching for POIs. That is, in searching for POIs, for example, POIs having names, etc. matching an input keyword are output and displayed ranked from the top of the popularity score. In this ranking, the popularity score is referenced.

[0013] Each of the functional units 11 to 15 included in the popularity estimation device 10 is configured to be able to access the various storage units 20, 30, 40, and 50. The popularity estimation device 10 and the various storage units 20, 30, 40, and 50 may be configured as a single device, or each may be configured as a separate device.

[0014] 1, the popularity estimation device 10 functionally comprises an acquisition unit 11, a generation unit 12, a score calculation unit 13, an output unit 14, and a model generation unit 15. These functional units 11 to 15 may be configured in one device, or may be distributed across multiple devices.

[0015] The block diagram shown in FIG. 1 shows functional blocks. These functional blocks (components) are realized by any combination of at least one of hardware and software. The method of realizing each functional block is not particularly limited. That is, each functional block may be realized by using one device that is physically or logically combined, or may be realized by using two or more devices that are physically or logically separated and directly or indirectly connected (for example, by using wires, wirelessly, etc.). The functional block may be realized by combining the one device or the multiple devices with software.

[0016] Functions include, but are not limited to, judgment, determination, judgment, calculation, computation, processing, derivation, investigation, search, confirmation, reception, transmission, output, access, resolution, selection, selection, establishment, comparison, assumption, expectation, consideration, broadcasting, notifying, communicating, forwarding, configuring, reconfiguring, allocation, mapping, assignment, etc. For example, a functional block (component) that performs the function of transmission is called a transmitting unit or a transmitter. In either case, as described above, there is no particular limitation on the method of realization.

[0017] For example, the popularity estimation device 10 according to an embodiment of the present invention may function as a computer. Fig. 2 is a diagram showing an example of a hardware configuration of the popularity estimation device 10 according to the present embodiment. The popularity estimation device 10 may be physically configured as a computer device including a processor 1001, a memory 1002, a storage 1003, a communication device 1004, an input device 1005, an output device 1006, a bus 1007, and the like.

[0018] In the following description, the term "apparatus" may be replaced with a circuit, a device, a unit, etc. The hardware configuration of the popularity estimation device 10 may be configured to include one or more of the devices shown in the drawings, or may be configured to exclude some of the devices.

[0019] Each function in the popularity estimation device 10 is realized by loading a specific software (program) onto hardware such as a processor 1001 and a memory 1002, so that the processor 1001 performs calculations and controls communication via a communication device 1004 and the reading and / or writing of data in the memory 1002 and storage 1003.

[0020] The processor 1001 controls the entire computer by running, for example, an operating system. The processor 1001 may be configured as a central processing unit (CPU) including an interface with peripheral devices, a control device, an arithmetic unit, a register, etc. For example, each of the functional units 11 to 16 shown in FIG. 1 may be realized by the processor 1001.

[0021] Moreover, the processor 1001 reads out a program (program code), software modules, and data from the storage 1003 and / or the communication device 1004 to the memory 1002, and executes various processes according to the programs. The programs used are those that cause a computer to execute at least a part of the operations described in the above-mentioned embodiments. For example, the functional units 11 to 15 of the popularity estimation device 10 may be realized by a control program stored in the memory 1002 and operated by the processor 1001. Although the above-mentioned various processes have been described as being executed by one processor 1001, they may be executed simultaneously or sequentially by two or more processors 1001. The processor 1001 may be implemented by one or more chips. The programs may be transmitted from a network via a telecommunication line.

[0022] The memory 1002 is a computer-readable recording medium, and may be composed of at least one of, for example, a read only memory (ROM), an erasable programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a random access memory (RAM), etc. The memory 1002 may be called a register, a cache, a main memory (primary storage device), etc. The memory 1002 can store a program (program code), a software module, etc. that can be executed to implement a popularity estimation method according to one embodiment of the present invention.

[0023] Storage 1003 is a computer-readable recording medium, and may be composed of at least one of, for example, an optical disk such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, a magneto-optical disk (e.g., a compact disk, a digital versatile disk, a Blu-ray® disk), a smart card, a flash memory (e.g., a card, a stick, a key drive), a floppy disk, a magnetic strip, and the like. Storage 1003 may be referred to as an auxiliary storage device. The above-mentioned storage medium may be, for example, a database, a server, or other suitable medium including memory 1002 and / or storage 1003.

[0024] The communication device 1004 is hardware (transmission / reception device) for performing communication between computers via a wired and / or wireless network, and is also called, for example, a network device, a network controller, a network card, or a communication module.

[0025] The input device 1005 is an input device (e.g., a keyboard, a mouse, a microphone, a switch, a button, a sensor, etc.) that accepts input from the outside. The output device 1006 is an output device (e.g., a display, a speaker, an LED lamp, etc.) that performs output to the outside. The input device 1005 and the output device 1006 may be integrated into one device (e.g., a touch panel).

[0026] Furthermore, each device such as the processor 1001 and the memory 1002 is connected by a bus 1007 for communicating information. The bus 1007 may be configured as a single bus, or may be configured as different buses between the devices.

[0027] The popularity estimation device 10 may be configured to include hardware such as a microprocessor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA), and some or all of the functional blocks may be realized by the hardware. For example, the processor 1001 may be implemented by at least one of these pieces of hardware.

[0028] Next, each functional unit of the popularity estimation device 10 will be described. The acquisition unit 11 acquires a first map image including a first POI (specific point) that is a target of popularity estimation. The acquisition unit 11 also acquires a plurality of second map images including a second POI that is a POI with a known popularity score indicating its popularity. Specifically, the acquisition unit 11 acquires a map image including the POI based on the location information of the POI.

[0029] Fig. 3 is a diagram showing an example of the structure of POI information stored in the POI information storage unit 20 and the data contained therein. Fig. 3(a) shows an example of POI information for a POI whose popularity score is known. Fig. 3(b) shows an example of POI information for a POI whose popularity score is unknown and for which a popularity score is to be estimated. As shown in Figs. 3(a) and 3(b), the POI information includes location information associated with each POI name that identifies the POI, population average (daytime population), distance to the nearest station, category ID, prefecture code, number of surrounding POIs, and popularity score.

[0030] The location information includes, for example, latitude and longitude information. The population average is, for example, the daytime population of a specified area (for example, an area of ​​several hundred meters square or an area with a radius of several hundred meters) centered on the location indicated in the location information, and is acquired, for example, based on the presence information of a mobile terminal. The category ID identifies the type to which the facility indicated by the POI belongs. The prefecture code identifies the prefecture to which the POI belongs. The number of surrounding POIs is the number of facilities of a specified category contained in a specified area centered on the location indicated in the location information.

[0031] The popularity score is a score indicating the popularity of the POI. The popularity score is generated based on, for example, the number of times the POI is mentioned on SNS etc. The popularity score of the POI shown in FIG. 3(b) is unknown because, for example, the POI is newly established and there is no mention history on SNS etc.

[0032] The map image storage unit 30 is configured with a storage device that stores a map image. The acquisition unit 11 acquires a map image by referring to the map image storage unit 30 based on the location information. The map image is an image of a map of an area, for example, 250 m square or 500 m square, centered on the location indicated in the location information. Note that the size of the area of ​​the map image is not limited to the size illustrated in the example.

[0033] The map image may be configured with multiple layers, i.e., the map image may include multiple layers such as topographical information, genre information, road maps, etc., relating to the same area.

[0034] The generation unit 12 inputs a map image into a trained geographical feature generation model to generate a feature vector representing the geographical features of the map image. The geographical feature generation model is a model constructed by machine learning that receives a map image as input and outputs a feature vector representing the geographical features of the map image.

[0035] Fig. 4 is a diagram showing a schematic example of generation of a feature vector indicating the geographical features of a POI. As shown in Fig. 4, the acquisition unit 11 acquires a first map image rp1 including a first POI from the map image storage unit 30. The first map image rp1 includes multiple hierarchical layers, such as topographical information rp11, genre information rp12, and a road map rp13. The acquisition unit 11 also acquires a second map image rp2 including a second POI from the map image storage unit 30. The second map image rp2 includes multiple hierarchical layers, such as topographical information rp21, genre information rp22, and a road map rp23.

[0036] The geographical feature generation model md1 is a model including a neural network. The geographical feature generation model md1 of this embodiment includes a convolution neural network (CNN).

[0037] The generation unit 12 inputs the first map image rp1 into the geographical feature generation model md1 to generate a first feature vector v1. The first feature vector v1 is a distributed representation of the first map image rp1 and represents the geographical features of the first POI.

[0038] Furthermore, the generation unit 12 inputs a plurality of second map images rp2 to the geographical feature generation model md1 to generate a second feature vector v2, which is a distributed representation of each of the second map images rp2.

[0039] Here, generation of the geographical feature generation model md1 will be described. The model generation unit 15 generates the geographical feature generation model md1 by machine learning. FIG. 5 is a diagram showing a functional configuration of the model generation unit 15. FIG. 6 is a diagram for explaining the learning process of the geographical feature generation model md1. As shown in FIG. 5, the model generation unit 15 includes a learning data acquisition unit 151, a vector generation unit 152, and a model learning unit 153.

[0040] The learning data acquisition unit 151 acquires reference learning data rpa, positive example learning data rpp, and negative example learning data rpn as a set of learning data to be used for learning the geographical feature generation model md1.

[0041] The reference learning data rpa includes a map image of a first area that serves as a reference in one set of learning data. The learning data acquisition unit 151 can acquire the map image of the first area from the map image storage unit 30 as the reference learning data rpa.

[0042] The positive example training data rpp includes a map image of the second area. In order to configure the training data as a positive example, the geographical features of the second area must be close to the geographical features of the first area. Therefore, the map image of the second area may be a rotated map image of the first area.

[0043] The negative example training data rpn includes a map image of a third region. To configure the training data as a negative example, the third region is set randomly, regardless of the first region.

[0044] The learning data acquisition unit 151 acquires map images of the first area, the second area, and the third area from the map image storage unit 30 as reference learning data rpa, positive example learning data rpp, and negative example learning data rpn. In addition, various attribute values ​​for each area may be associated with the map image. Specifically, the daytime population for each area, the number of facilities (number of POIs) for each category, and the like may be associated with the map image. Therefore, the learning data acquisition unit 151 can acquire various attribute values ​​associated with the first area, the second area, and the third area by including them in the reference learning data rpa, the positive example learning data rpp, and the negative example learning data rpn.

[0045] Furthermore, the learning data acquiring unit 151 can extract the number of POIs belonging to each category for each of the first area, the second area, and the third area based on the category ID and the location information by referring to the POI information storage unit 20. Then, the learning data acquiring unit 151 can include the extracted number of POIs in the reference learning data rpa, the positive example learning data rpp, and the negative example learning data rpn as attribute values ​​of each area.

[0046] Since the learning data acquisition unit 151 can acquire attribute values ​​for each geographical region as described above, the learning data acquisition unit 151 may acquire a map image of a second region having an attribute value whose difference from a predetermined attribute value for the first region is equal to or smaller than a predetermined value as the positive example learning data rpp. The learning data acquisition unit 151 may acquire a map image of a third region having an attribute value whose difference from a predetermined attribute value for the first region exceeds a predetermined value as the negative example learning data rpn.

[0047] The vector generation unit 152 inputs the reference training data rpa, the positive example training data rpp, and the negative example training data rpn to the geographical feature generation model md1 to generate a reference feature vector va, a positive example feature vector vp, and a negative example feature vector vn. That is, the reference feature vector va, the positive example feature vector vp, and the negative example feature vector vn are distributed representations of the reference training data rpa, the positive example training data rpp, and the negative example training data rpn. As shown in Fig. 6, the geographical feature generation models md1 to which the reference training data rpa, the positive example training data rpp, and the negative example training data rpn are input share parameters.

[0048] The model learning unit 153 adjusts parameters of the neural network (CNN) included in the geographical feature generation model md1 so that the difference between the reference feature vector va and the positive example feature vector vp approaches zero and the difference between the reference feature vector va and the negative example feature vector vn becomes larger.

[0049] Specifically, the model learning unit 153 defines the distance dp between the reference feature vector va and the positive example feature vector vp as Δ(a,p)=||ap||2, and defines the distance dn between the reference feature vector va and the negative example feature vector vn as Δ(a,n)=||an||2. The distances dp and dn may be Euclidean distances.

[0050] Then, the model learning unit 153 updates the parameters of the neural network (CNN) included in the geographical feature generation model md1 so that the distance dp approaches zero and the distance dn becomes larger. That is, the model learning unit 153 updates the parameters of the neural network (CNN) included in the geographical feature generation model md1 so that the distance dp approaches zero and the distance dn becomes larger. + ,d - ) = SoftMax(Δ(a,p),Δ(a,n)), and the loss function Loss(d + ,d - The CNN parameters are updated (learned) using machine learning to bring σ(σ) closer to zero.

[0051] The generation unit 12 uses the geographical feature generation model md1 thus generated by the model generation unit 15 to generate a feature vector that is a distributed representation of a map image of a predetermined area including a POI.

[0052] Referring again to Figure 1, the score calculation unit 13 calculates the popularity score of the first POI from the popularity score of the second POI based on the degree of similarity between a first feature vector representing the geographical characteristics of the first POI and a second feature vector representing the geographical characteristics of the second POI.

[0053] Specifically, the score calculation unit 13 calculates the popularity score of the first POI from the popularity score of the second POI based on the Euclidean distance between the first feature vector and the second feature vector.

[0054] More specifically, the score calculation unit 13 extracts a top predetermined number of second POIs from among the multiple second POIs in order of the Euclidean distance between the second feature vector of each second POI and the first feature vector. Then, the score calculation unit 13 calculates the popularity score of the first POI based on the popularity scores of the multiple extracted second POIs and the Euclidean distance from the first POI. For example, the score calculation unit 13 may obtain the popularity score of the first POI by weighting the popularity scores of the multiple extracted second POIs based on the Euclidean distance from the first POI and calculating a weighted average.

[0055] Since the Euclidean distance between feature vectors reflects the geographical similarity between POIs, calculating the popularity score of the first POI in this manner makes it possible to estimate the popularity score of the first POI that appropriately reflects the geographical similarity with the second POI. Also, by using the popularity scores of a predetermined number of second POIs ranked in order of closest Euclidean distance to the first feature vector, it becomes possible to calculate the popularity score of the first POI with higher accuracy.

[0056] In addition, the score calculation unit 13 may estimate the popularity score of the first POI based on the first feature vector of the first POI by using a popularity estimation model constructed by machine learning using input data including at least a second feature vector related to the second POI and learning data including ground truth labels consisting of the popularity scores of each second POI.

[0057] Fig. 7 is a diagram showing a model of the learning process of the popularity estimation model md2. As shown in Fig. 7, the popularity estimation model md2 is a model including a neural network constructed by machine learning based on learning data in which a second feature vector v2 representing a geographic feature of a second POI whose popularity score is known is used as input data (feature amount) and the popularity score sc2 of the second POI is used as a correct answer label.

[0058] The popularity estimation model md2, which is a model including a trained neural network, can be considered as a program that is loaded or referenced by a computer and causes the computer to execute specified processes and realize specified functions.

[0059] That is, the trained popularity estimation model md2 of this embodiment is used in a computer equipped with a CPU and a memory. Specifically, the CPU of the computer operates in accordance with instructions from the trained popularity estimation model md2 stored in the memory to perform calculations based on trained weighting coefficients and response functions corresponding to each layer for input data (feature vectors representing geographical features of POIs, etc.) input to the input layer of the neural network, and to output results (load and power generation) from the output layer.

[0060] The neural network included in the popularity estimation model md2 may be, for example, a feedforward neural network, a convolutional neural network, etc., and the type of neural network used is not limited.

[0061] Furthermore, the input data of the learning data used to train the popularity estimation model md2 may include predetermined attribute information related to the second POI. The predetermined attribute information may be, for example, the population, the number of facilities in a specific category, the distance to the nearest station, etc.

[0062] In this case, the score calculation unit 13 inputs the first feature vector of the first POI and predetermined attribute information related to the first POI into the popularity estimation model md2 to estimate the popularity score of the first POI.

[0063] The popularity estimation model md2 constructed in this manner calculates the popularity score of the first POI, making it possible to estimate the popularity score of the first POI that appropriately reflects the geographical similarity with the second POI. In addition, if the input feature amount to the popularity estimation model md2 includes predetermined attribute information related to the POI, it becomes possible to estimate the popularity score of the first POI that reflects the attribute similarity as well as the geographical similarity with the second POI.

[0064] 1 again, the output unit 14 outputs the calculated popularity score of the first POI. Specifically, the output unit 14 stores the calculated popularity score in the popularity score field of the POI information of the first POI shown in, for example, FIG. 3(b).

[0065] Next, with reference to FIG. 8 and FIG. 9, a process of assigning the popularity score assigned to the set to which the POI belongs to, to each POI belonging to the set will be described.

[0066] In this example, the acquisition unit 11 acquires a specific location as the second POI, which belongs to a second set different from the first set to which the first POI belongs and is included in the same category as the first set, and whose popularity score is known, and acquires a map image including the second POI as the second map image.

[0067] FIG. 8 is a diagram showing an example of the configuration of POI information stored in the POI information storage unit 20 and the data included therein of POI information belonging to a specific set. FIG. 8(a) is a diagram showing an example of POI information of each shop (second POI) of store A, which belongs to store B (second set) different from store A, which is included in the same category (shop category) as store A (first set), and has a known popularity score. FIG. 8(b) is a diagram showing an example of POI information of each shop (first POI) of store A to which a popularity score is assigned. The POI information shown in FIG. 8 includes, like the POI information shown in FIG. 3, location information associated with each POI name that identifies the POI, population average (daytime population), distance to the nearest station, category ID, prefecture code, number of surrounding POIs, and popularity score.

[0068] The acquisition unit 11 refers to the map image storage unit 30 and acquires map images of each store (C1 to C3, ...) of Store A as a second map image based on the location information of each store. The acquisition unit 11 also refers to the map image storage unit 30 and acquires map images of each store (D1 to D3, ...) of Store B as a first map image based on the location information of each store. The generation unit 12 inputs the second map image and the first map image to a geographical feature generation model md1 to generate a second feature vector and a first feature vector.

[0069] The score calculation unit 13 uses a feature vector indicating the geographical features of each store of Store A as input data and learning data (teacher data) having the popularity score of each store of Store A as a correct answer label to construct a popularity estimation model md2 by machine learning similar to the model generation described with reference to Fig. 7. Then, the score calculation unit 13 acquires the popularity score of each store (the rightmost column in Fig. 8(b)) obtained by inputting a first feature vector indicating the geographical features of each store of Store B into the popularity estimation model md2 constructed using information about Store A as the teacher data, as the first popularity score of each store.

[0070] The output unit 14 pro rataly distributes the set popularity score assigned to the first set, indicating the popularity of the first set, among the multiple first POIs included in the first set, based on the first popularity scores assigned to each of the first POIs, and outputs the pro rata scores as second popularity scores for each of the first POIs.

[0071] FIG. 9 is a diagram showing an example of calculation of the second popularity score based on the first popularity score, and shows an example of calculation of the second popularity score of each shop of Store B.

[0072] As shown by the popularity score sc81 of each store in Store B in Figure 9, the score calculation unit 13 calculates the first popularity score (500, 50, 10, ...) for each store in Store B (Store D1, Store D2, Store D3, ...).

[0073] The first popularity score is estimated by the popularity estimation model md2 constructed using the geographical features and popularity scores of each store in Store A, and therefore reflects the relationship between the geographical features and popularity scores of each store in Store A. That is, for example, if a store in Store A has a high popularity score, the first popularity score of Store B, which has similar geographical features to that store, will be high.

[0074] Then, if the sum of the first popularity scores calculated for each shop in Store B is "25000" and the popularity score scB assigned to the entire set of Store B is "5000", the output unit 14 pro rataly distributes the popularity score "5000" for Store B based on the first popularity scores (500, 50, 100, ...) of each shop in Store B (Store D1, Store D2, Store D3, ...) and outputs it as the second popularity scores (100, 1, 2, ...) of each shop in Store B (Store D1, Store D2, Store D3, ...). For example, the second popularity score (100) of Store B's Store D1 is calculated by the following formula: Store B's popularity score 5000 x (Store D1's first popularity score 500 / total first popularity scores of all stores in Store B 25000)

[0075] In this way, a popularity estimation model is constructed using the feature vectors representing the geographical features and popularity scores of each store of Store A that is included in the same category as Store B as learning data, and the collective popularity score assigned to Store B is apportioned to each store of Store B based on the first popularity score of each store of Store B obtained using the constructed popularity estimation model, so that the collective popularity score assigned to the group can be appropriately assigned to each store.

[0076] FIG. 10 is a flowchart showing the process of the popularity estimation method in the popularity estimation device 10. As shown in FIG.

[0077] In step S1, the acquisition unit 11 acquires a first map image including a first POI that is a POI for which a popularity score is to be estimated, and also acquires a plurality of second map images including a second POI that is a POI for which a popularity score is known.

[0078] In step S2, the generation unit 12 inputs the first map image and the second map image to the trained geographical feature generation model md1 to generate a first feature vector and a second feature vector.

[0079] In step S3, the score calculation unit 13 calculates the popularity score of the first POI from the popularity score of the second POI based on the degree of similarity between the first feature vector and each of the plurality of second feature vectors.

[0080] In step S4, the output unit 14 outputs the popularity score of the first POI.

[0081] Next, a description will be given of a popularity estimation program for causing a computer to function as the popularity estimation device 10 of this embodiment. Fig. 11 is a diagram showing the configuration of the popularity estimation program.

[0082] The popularity estimation program P1 includes a main module m10 that controls the popularity estimation process in the popularity estimation device 10, an acquisition module m11, a generation module m12, a score calculation module m13, an output module m14, and a model generation module m15. The modules m11 to m15 realize the functions of the acquisition unit 11, the generation unit 12, the score calculation unit 13, the output unit 14, and the model generation unit 15.

[0083] The popularity estimation program P1 may be transmitted via a transmission medium such as a communication line, or may be stored in a recording medium M1 as shown in FIG.

[0084] In the popularity estimation device 10, the popularity estimation method, and the popularity estimation program P1 of the present embodiment described above, a first feature vector is obtained by inputting a first map image including a first specific point into a geographical feature generation model. The first feature vector represents the geographical features of the first specific point. Also, a second feature vector is obtained by inputting a second map image including a second specific point, the popularity score of which is known, into the geographical feature generation model. The second feature vector represents the geographical features of the second specific point. Then, based on the degree of similarity between the first feature vector and the second feature vector, the popularity score of the first specific point is calculated from the popularity score of the known second specific point, so that it is possible to obtain a popularity score of the first specific point that appropriately reflects the geographical similarity with the second specific point.

[0085] In addition, in a popularity estimation system according to another embodiment, the score calculation unit may calculate the popularity score of the first specific location from the popularity score of the second specific location based on the Euclidean distance between the first feature vector and a plurality of second feature vectors.

[0086] Since the Euclidean distance between feature vectors reflects the geographical similarity between specific locations, according to the above embodiment, the popularity score of the first specific location is calculated based on the Euclidean distance between the first feature vector and multiple of the second feature vectors, making it possible to estimate the popularity score of the first specific location that appropriately reflects the geographical similarity with the second specific location.

[0087] In addition, in a popularity estimation system according to another embodiment, the score calculation unit may extract a top predetermined number of second specific locations from among a plurality of second specific locations in order of the Euclidean distance between the second feature vector of each second specific location and the first feature vector, and calculate the popularity score of the first specific location based on the popularity scores and Euclidean distances of the extracted plurality of second specific locations.

[0088] According to the above embodiment, the popularity score of the first specific location is calculated based on the popularity scores of a top predetermined number of second specific locations in order of Euclidean distance to the first feature vector, so that the popularity score of the second specific location having a high degree of geographical similarity to the first specific location is used to calculate the popularity score of the first specific location. This makes it possible to estimate the popularity score of the first specific location with higher accuracy.

[0089] In addition, in another embodiment of the popularity estimation system, the score calculation unit may estimate the popularity score of the first specific location based on the first feature vector of the first specific location using a popularity estimation model constructed by machine learning using input data including at least a second feature vector related to the second specific location and learning data including ground truth labels consisting of the popularity scores of each second specific location.

[0090] According to the above embodiment, the popularity score of the first specific location is calculated by a popularity estimation model constructed by machine learning using learning data consisting of a pair of the second feature vector of the second specific location and the popularity score of the second specific location, making it possible to estimate the popularity score of the first specific location that appropriately reflects the geographical similarity with the second specific location.

[0091] In addition, in a popularity estimation system according to another embodiment, the input data of the learning data used to train the popularity estimation model may include specified attribute information relating to the second specific location, and the score calculation unit may input a first feature vector of the first specific location and the specified attribute information relating to the first specific location into the popularity estimation model to estimate the popularity score of the first specific location.

[0092] According to the above embodiment, it is possible to estimate the popularity score of the first specific location that reflects not only the geographical similarity with the second specific location but also the similarity in attributes.

[0093] In addition, in another embodiment of the popularity estimation system, the acquisition unit acquires a map image including the second specific point as the second map image, the second specific point being a specific point that belongs to a second set different from the first set included in the same category as the first set to which the first specific point belongs and has a known popularity score, the score calculation unit acquires a popularity score estimated based on the first feature vector of the first specific point as the first popularity score using a popularity estimation model constructed by machine learning using learning data including input data including a second feature vector based on the second map image acquired by the acquisition unit and learning data including a ground truth label consisting of the popularity score of each second specific point, and the output unit may output a score obtained by dividing the known score assigned to the first set, the set popularity score indicating the popularity of the set, based on the first popularity score of each first specific point, for a plurality of first specific points included in the first set, as the second popularity score of each first specific point.

[0094] According to the above embodiment, a popularity estimation model is used that appropriately reflects the relationship between the feature vector and the popularity score in the category that includes the first set and the second set. Then, based on the first popularity score acquired using such a popularity estimation model, the set popularity score assigned to the first set to which the first specific point belongs is pro rataly distributed to each of the first specific points, so that the popularity score assigned to the set to which the specific point belongs can be appropriately assigned to each specific point.

[0095] In addition, a popularity estimation system according to another embodiment may further include a model generation unit that receives an input of a map image of a specific location, outputs a feature vector representing a geographical feature of the map image, and generates a geographical feature generation model including a neural network, wherein the model generation unit may include a learning data acquisition unit that acquires reference learning data including a map image of a first region, positive example learning data including a map image of a second region having an attribute value whose difference from a predetermined attribute value related to the first region is equal to or smaller than a predetermined value, and negative example learning data including a map image of a third region having an attribute value whose difference from the predetermined attribute value related to the first region exceeds a predetermined value, a vector generation unit that inputs each of the reference learning data, the positive example learning data, and the negative example learning data into the geographical feature generation model to generate a reference feature vector, a positive example feature vector, and a negative example feature vector, and a model learning unit that adjusts parameters of the neural network so that a difference between the reference feature vector and the positive example feature vector approaches zero and a difference between the reference feature vector and the negative example feature vector becomes larger.

[0096] According to the above aspect, positive example training data including a map image of a second region having a high degree of similarity to the first region in geographical features, and negative example training data including a map image of a third region having a low degree of similarity to the first region in geographical features are obtained. Then, by inputting the reference training data, the positive example training data, and the negative example training data into a geographical feature generation model, a reference feature vector, a positive example feature vector, and a negative example feature vector are obtained, and the parameters of the neural network are adjusted so that the difference between the reference feature vector and the positive example feature vector approaches zero and the difference between the reference feature vector and the negative example feature vector becomes larger, so that a geographical feature generation model that outputs a feature vector in which the degree of similarity of geographical features is appropriately reflected can be obtained.

[0097] In addition, in another aspect of the popularity estimation system, a computer is caused to function so as to input a map image of a specific geographically identified point and output a feature vector representing the geographical features of the map image, and a trained geographical feature generation model including a neural network is constructed by inputting reference training data including a map image of a first region, positive example training data including a map image of a second region having an attribute value whose difference from a predetermined attribute value related to the first region is equal to or less than a predetermined value, and negative example training data including a map image of a third region having an attribute value whose difference from the predetermined attribute value related to the first region exceeds a predetermined value, to the geographical feature generation model to generate a reference feature vector, a positive example feature vector, and a negative example feature vector, and adjusting parameters of the neural network by machine learning so that the difference between the reference feature vector and the positive example feature vector approaches zero and the difference between the reference feature vector and the negative example feature vector becomes larger.

[0098] According to the above aspect, it is possible to obtain a geographical feature generation model that outputs a feature vector in which the degree of similarity between geographical features is appropriately reflected.

[0099] Although the present embodiment has been described in detail above, it is clear to those skilled in the art that the present embodiment is not limited to the embodiment described in this specification. The present embodiment can be implemented as a modified or altered form without departing from the spirit and scope of the present invention defined by the description of the claims. Therefore, the description in this specification is intended to be illustrative and does not have any limiting meaning on the present embodiment.

[0100] Each aspect / embodiment described herein may be applied to systems utilizing LTE (Long Term Evolution), LTE-Advanced (LTE-A), SUPER 3G, IMT-Advanced, 4G, 5G, FRA (Future Radio Access), W-CDMA (registered trademark), GSM (registered trademark), CDMA2000, UMB (Ultra Mobile Broadband), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, UWB (Ultra-WideBand), Bluetooth (registered trademark), or other suitable systems and / or next generation systems enhanced thereon.

[0101] The steps, sequences, flow charts, etc. of each aspect / embodiment described herein may be reordered unless inconsistent, For example, the methods described herein present elements of various steps in an example order, and are not limited to the particular order presented.

[0102] The input and output information may be stored in a specific location (e.g., memory) or may be managed in a management table. The input and output information may be overwritten, updated, or added to. The output information may be deleted. The input information may be transmitted to another device.

[0103] The determination may be based on a value represented by a single bit (0 or 1), a Boolean (true or false) value, or a numerical comparison (e.g., comparison with a predetermined value).

[0104] Each aspect / embodiment described in this specification may be used alone, in combination, or switched according to execution. In addition, notification of specific information (e.g., notification that "X is the case") is not limited to being done explicitly, but may be done implicitly (e.g., not notifying the specific information).

[0105] Although the present disclosure has been described in detail above, it is clear to those skilled in the art that the present disclosure is not limited to the embodiments described in the present disclosure. The present disclosure can be implemented in modified and altered forms without departing from the spirit and scope of the present disclosure as defined by the claims. Therefore, the description of the present disclosure is intended to be illustrative and does not have any limiting meaning on the present disclosure.

[0106] Software shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executable files, threads of execution, procedures, functions, etc., whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.

[0107] Additionally, software, instructions, etc. may be transmitted or received over a transmission medium. For example, if the software is transmitted from a website, server, or other remote source using wired technologies, such as coaxial cable, fiber optic cable, twisted pair, and Digital Subscriber Line (DSL), and / or wireless technologies, such as infrared, radio, and microwave, these wired and / or wireless technologies are included within the definition of transmission media.

[0108] The information, signals, etc. described in this disclosure may be represented using any of a variety of different technologies. For example, the data, instructions, commands, information, signals, bits, symbols, chips, etc. that may be referred to throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or magnetic particles, optical fields or photons, or any combination thereof.

[0109] In addition, terms explained in this disclosure and / or terms necessary for understanding this specification may be replaced with terms having the same or similar meanings.

[0110] As used herein, the terms "system" and "network" are used interchangeably.

[0111] Furthermore, the information, parameters, etc. described in this specification may be expressed as absolute values, may be expressed as relative values ​​from a predetermined value, or may be expressed as other corresponding information.

[0112] As used in this disclosure, the terms "determining" and "determining" may encompass a wide variety of actions. "Determining" and "determining" may include, for example, judging, calculating, computing, processing, deriving, investigating, looking up, search, inquiry (e.g., searching in a table, database, or other data structure), ascertaining, and the like. "Determining" and "determining" may also include receiving (e.g., receiving information), transmitting (e.g., sending information), input, output, accessing (e.g., accessing data in a memory), and the like. In addition, "judgment" and "decision" can include considering resolving, selecting, choosing, establishing, comparing, etc., to be a "judgment" or "decision." In other words, "judgment" and "decision" can include considering some action to be a "judgment" or "decision." In addition, "judgment" can be interpreted as "assuming," "expecting," "considering," etc.

[0113] As used in this disclosure, the phrase "based on" does not mean "based only on," unless expressly stated otherwise. In other words, the phrase "based on" means both "based only on" and "based at least on."

[0114] When designations such as "first," "second," and the like are used herein, any reference to an element is not intended to generally limit the quantity or order of those elements. These designations may be used herein as a convenient method of distinguishing between two or more elements. Thus, a reference to a first and a second element does not imply that only two elements may be employed therein or that the first element must precede the second element in some way.

[0115] To the extent that the terms "include," "including," and variations thereof are used herein in the specification or claims, these terms are intended to be inclusive, similar to the term "comprising." Further, the term "or" as used herein is not intended to be an exclusive or.

[0116] In this specification, a plurality of devices is also included unless the context or technical basis clearly indicates that only one device is present.

[0117] Throughout this disclosure, the plural is intended to be included unless the context clearly indicates the singular. [Explanation of symbols]

[0118] DESCRIPTION OF THE PREFERRED EMBODIMENTS 1...Popularity estimation system, 10...Popularity estimation device, 11...Acquisition unit, 12...Generation unit, 13...Score calculation unit, 14...Output unit, 15...Model generation unit, 20...POI information storage unit, 30...Map image storage unit, 40...Geographical feature generation model storage unit, 50...Popularity estimation model storage unit, 151...Learning data acquisition unit, 152...Vector generation unit, 153...Model learning unit, M1...Recording medium, m10...Main module, m11...Acquisition module, m12...Generation module, m13...Score calculation module, m14...Output module, m15...Model generation module, P1...Popularity estimation program.

Claims

1. A popularity estimation system for estimating the popularity of a specific geographically specified location, comprising: an acquisition unit that acquires a first map image including a first specific point that is a specific point to be estimated, and acquires a plurality of second map images including a second specific point that is a specific point having a known popularity score indicating a popularity; a generation unit that inputs the first map image and the second map image into a trained geographic feature generation model that receives a map image as an input and outputs a feature vector representing a geographic feature of the map image, and generates a first feature vector and a second feature vector representing the geographic features of the first map image and the second map image, respectively; a score calculation unit that calculates a popularity score of the first specific location from a popularity score of the second specific location based on a degree of similarity between the first feature vector and the second feature vector; an output unit that outputs the calculated popularity score of the first specific location; A popularity estimation system comprising:

2. the score calculation unit calculates a popularity score of the first specific location from a popularity score of the second specific location based on a Euclidean distance between a first feature vector and a plurality of the second feature vectors; The popularity estimation system according to claim 1 .

3. the score calculation unit extracts a top predetermined number of second specific locations from among the plurality of second specific locations in order of the Euclidean distance between the second feature vector of each second specific location and the first feature vector, and calculates a popularity score of the first specific location based on the popularity scores and Euclidean distances of the extracted plurality of second specific locations; The popularity estimation system according to claim 2 .

4. the score calculation unit estimates a popularity score of the first specific location based on the first feature vector of the first specific location by using a popularity estimation model constructed by machine learning using input data including at least the second feature vector related to the second specific location and learning data including a ground truth label consisting of a popularity score of each second specific location; The popularity estimation system according to claim 1 .

5. the input data of the learning data used for learning the popularity estimation model includes predetermined attribute information related to a second specific location; the score calculation unit inputs the first feature vector of the first specific location and the predetermined attribute information related to the first specific location into the popularity estimation model to estimate a popularity score of the first specific location; The popularity estimation system according to claim 4 .

6. the acquisition unit acquires, as the second map image, a map image including a second specific point, the second specific point being a specific point that belongs to a second set that is different from the first set and that is included in the same category as a first set to which the first specific point belongs, and whose popularity score is known; the score calculation unit acquires, as a first popularity score, a popularity estimation model constructed by machine learning using input data including a second feature vector based on the second map image acquired by the acquisition unit and learning data including a ground truth label consisting of a popularity score of each second specific location, the popularity score estimated based on the first feature vector of the first specific location; the output unit outputs a set popularity score, which is a known score assigned to the first set and indicates the popularity of the set, as a second popularity score for each of the first specific locations included in the first set, the set popularity score being proportionally divided based on the first popularity score of each of the first specific locations; The popularity estimation system according to claim 1 .

7. a model generation unit that receives a map image of a specific location as an input, outputs a feature vector representing geographical features of the map image, and generates a geographical feature generation model including a neural network; The model generation unit a learning data acquisition unit that acquires reference learning data including a map image of a first region, positive example learning data including a map image of a second region having an attribute value whose difference from a predetermined attribute value related to the first region is equal to or smaller than a predetermined value, and negative example learning data including a map image of a third region having an attribute value whose difference from the predetermined attribute value related to the first region exceeds a predetermined value; a vector generation unit that inputs the reference training data, the positive example training data, and the negative example training data into the geographical feature generation model to generate a reference feature vector, a positive example feature vector, and a negative example feature vector; and a model learning unit that adjusts parameters of the neural network so that a difference between the reference feature vector and the positive example feature vector approaches zero and a difference between the reference feature vector and the negative example feature vector becomes larger. The popularity estimation system according to any one of claims 1 to 6.