Estimation device
The estimation device uses deep learning and graph neural networks to construct models for both user groups' data, accurately estimating missing values and enhancing data integration by addressing the challenge of missing data in integrated user group data, ensuring high accuracy and utility.
Patent Information
- Application Number
- PCT/JP2024/019094
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-23
- Filing Date
- 2024-05-23
- Publication Date
- 2024-11-28
AI Technical Summary
When integrating data from two user groups with some common users, missing data in the first and second table data can lead to a decrease in utility value, especially for users who are not common, making it difficult to obtain integrated data accurately.
An estimation device with a first learning section to construct a first estimation model for the first table data and a second learning section to construct a second estimation model for the second table data, using deep learning technology and graph neural networks to estimate missing data, and an estimation unit to combine these models to accurately estimate missing data in both tables, including for users not in the common user group.
The device effectively complements missing data with high accuracy, enhancing the utility of integrated data by leveraging the attributes and behavior information of both user groups, reducing bias and improving estimation precision.
Smart Images

Figure JP2024019094_28112024_PF_FP_ABST
Abstract
Description
estimation device
[0001] The present disclosure relates to an estimation device.
[0002] Patent Literature 1 discloses a technology related to data imputation. This technology calculates the degree of correlation with other data items when missing data values exist in multiple data records, and performs imputation of the data values using a regression method based on the data item values.
[0003] Japanese Patent Application Laid-Open No. 2020-154828
[0004] Consider integrating first table data storing first data for the first user group and second user group, some of which are common users, with second table data storing second data for the second user group. If the first table data and second table data contain missing data, the usefulness of the integrated data may be reduced if the missing data remains in the integrated data. Furthermore, if the first data and second data have different data items, it is possible to obtain integrated data for the common users because the first data and second data for the common users exist. However, for example, it is difficult to obtain integrated data for such users because there is no second data for the first users other than the common users.
[0005] The present disclosure has been made in view of the above, and aims to provide an estimation device that can complement data with high accuracy.
[0006] An estimation device according to the present disclosure includes a first learning unit, a second learning unit, and an estimation unit. The first learning unit constructs a first estimation model. The first estimation model estimates missing data in the first table data based on first table data configured with first data indicating attributes of each user in a first user group. The second learning unit constructs a second estimation model. The second estimation model estimates missing data of a common user group in the second table data based on the second table data and first data about the common user group in the first table data. The second table data includes second data indicating attributes of each user in a second user group, which is a common user group common to a portion of the first user group. The estimation unit estimates values of the second data for the first user group excluding the common user group, based on the first estimation model and the second estimation model.
[0007] In the above estimation device, missing data in the first table data is complemented by the first estimation model, and missing data in the second table data is complemented by the second estimation model. Second data for a first user other than a common user is complemented based on the first estimation model and the second estimation model. The second estimation model, which estimates the missing data in the second table data, is constructed based on the first data and the second data and reflects the contents of the first data. Therefore, the estimation unit, which estimates data based on the first estimation model and the second estimation model, can accurately estimate second data that is not originally included in the first table data.
[0008] According to the present disclosure, data can be complemented with high accuracy.
[0009] FIG. 1 is a diagram illustrating a configuration of an example estimation device according to an embodiment of the present disclosure. FIG. 2 is a diagram illustrating the contents of example first table data. FIG. 3 is a diagram illustrating the contents of example second table data. FIG. 4 is a diagram schematically illustrating an example first estimation model. FIG. 5 is a diagram illustrating the contents of example integrated data. FIG. 6 is a diagram schematically illustrating an example second estimation model. FIG. 7 is a diagram illustrating the hardware configuration of an example estimation device.
[0010] Hereinafter, an embodiment of an estimation device according to the present disclosure will be described in detail with reference to the drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and duplicated description will be omitted.
[0011] FIG. 1 is a diagram illustrating the configuration of an example estimation device 10. The example estimation device 10 may be used in a system that complements missing data, etc., in the integrated table data when integrating two sets of table data having different attribute data. An example of integrating first table data and second table data will be described below. The first table data is composed of information about a first user group that uses a first service. The second table data is composed of information about a second user group that uses a second service. In one example, the first service and the second service may be services that award points to users in response to their purchases, etc., but the content of the first service and the second service is not limited thereto. The first user group that uses the first service and the second user group that uses the second service share some common users. That is, the first user group includes users who also use the second service (common users). In this embodiment, for simplicity of explanation, it is assumed that all members of the second user group are common users. For example, in a case where the company that operates the first service is the company itself and the company that operates the second service is a partner company, when the partner company provides second table data for only common users, the estimation device 10 is used by the company itself.
[0012] As shown in FIG. 1 , an example estimation device 10 includes an input unit 11, a preprocessing unit 12, a first learning unit 13, a second learning unit 15, an estimation unit 16, and an output unit 17. The input unit 11 includes a first data input unit 11A and a second data input unit 11B. The first data input unit 11A receives the contents of first table data, which stores data (first data) of a first group of users who use a first service. The first table data may be managed by the company. The second data input unit 11B receives the contents of data about a common user group in the second table data. In this embodiment, since the second group of users constituting the second table data are all common users, the contents of the second table data may be input directly to the second data input unit 11B. The second table data may be managed by a partner company. For example, the second data may be data about common users extracted from the second table data managed by the partner company.
[0013] FIG. 2 is a diagram illustrating the contents of the first table data. FIG. 3 is a diagram illustrating the contents of the second table data. An example of the first data is composed of a user ID (identifier) for identifying a user and attribute data for each user. The user attribute data may include basic information and behavioral information. The basic information may include static characteristics such as the user's gender, age, and place of residence. The basic information may be based on contract information acquired from the user when the user begins using the first service. The behavioral information includes dynamic characteristics resulting from the user's behavior, such as the user's payment amount related to the first service. The behavioral information may include multiple behavioral histories with different characteristics. In the example of FIG. 2, the age field for the user with user ID 2 and the payment amount field for the user with user ID 4 are missing data, indicated by "-".
[0014] The user's behavior information may be acquired via an application running on a terminal device. The terminal device is, for example, a communication device operated by a user. The terminal device may be, for example, a mobile terminal such as a highly functional mobile phone (smartphone), a mobile phone, or a personal digital assistant (PDA). The devices constituting the terminal device are not limited to these.
[0015] The second data, as an example, is composed of a common user ID for identifying a common user and data on the attributes of each user. The common user ID may have the same identification code as the user ID in the first data. The user ID (common user ID) may also be personally identifiable information such as an email address or a mobile phone number. In this case, users with a common user ID included in both the first table data and the second table data can be extracted as common users. In the example shown in FIG. 3 , users with user IDs 2, 3, 4, and 7 from the first user group constituting the first table data are illustrated as common users. Also, in the example shown in FIG. 3 , the favorite store field for the user with common user ID 3, the residential area and favorite store field for the user with user ID 4, and the residential area field for the user with user ID 7 are all missing data, indicated by "-".
[0016] In the second data, the user attributes may include basic information and behavioral information. The basic information includes static characteristics such as the user's residential area. In addition to the residential area, the basic information may also include gender, as in the first data. In this example, basic information included in the second table data that overlaps with the first data may be deleted before being input into the second data input unit 11B. The behavioral information includes dynamic characteristics resulting from the user's behavior in the second service, such as the user's membership rank, usage points, and favorite stores.
[0017] The preprocessing unit 12 performs a process of converting the first data input to the first data input unit 11A and the second data input to the second data input unit 11B into a format that can be processed by the first learning unit 13 and the second learning unit 15. In one example, the preprocessing unit 12 may convert character string data included in the first data and the second data into numerical data.
[0018] The preprocessing unit 12 also represents the input table data in a graph space. FIG. 4 schematically illustrates an example of first table data represented in a graph space as an example of table data represented in a graph space. In this embodiment, the table data is composed of a user ID identifying each user and a value for each item in a data string associated with the user ID. Therefore, the preprocessing unit 12 represents the user and each item in the data string as a node. In FIG. 4, each user ID is represented as a node 31 surrounded by a circle, and each item is represented as a node 32 surrounded by a square. The preprocessing unit 12 connects the corresponding user node 31 and item node 32 using the item value as an edge 33. The preprocessing unit 12 embeds the nodes 31, 32, and edges 33 so that the user, item, and value are represented as vectors. Hereinafter, the user vector 31a generated by the embedded representation will be referred to as a feature vector, the item vector 32a will be referred to as a variable vector, and the value vector 33a will be referred to as a weight vector.
[0019] The preprocessing unit 12 represents the first table data in a graph space as preprocessing for model construction in the first learning unit 13. Furthermore, the preprocessing unit 12 represents table data (common table data) in which the first data and second data for a common user group are integrated in a graph space as preprocessing for model construction in the second learning unit 15.
[0020] For example, the preprocessing unit 12 may generate common table data by integrating the first data of the common user group constituting the first table data and the second data of the common user group constituting the second table data, and represent the generated common table data in a graph space. FIG. 5 is a diagram showing the contents of the common table data. An example of the common table data is composed of a user ID and data on the attributes of each user. The user attributes include basic information and behavioral information in the first data and basic information and behavioral information in the second data. For ease of understanding, FIG. 5 directly integrates the data related to the common users in the first data shown in FIG. 2 and the second data shown in FIG. 3 . However, data converted into numerical data by the preprocessing unit 12 may be used for each data.
[0021] The first learning unit 13 constructs a first estimation model using a deep learning technique with a neural network (here, a graph neural network) and the first table data in the graph space processed by the preprocessing unit 12 as learning data. For example, the first learning unit 13 constructs a learning model that estimates the value of each item from the feature vector and the variable vector by repeating learning using data for each user while updating the initial values assigned to the feature vector and the variable vector in the preprocessing unit 12, based on an adjacency matrix that represents the positional relationship between the user node and the item node, a feature vector that represents the user's characteristics, a variable vector that represents the item's characteristics, and a weight vector that represents the value's characteristics. That is, the first estimation model trained by the first learning unit 13 has a feature vector and a variable vector for estimating the value of each item in the first table data.
[0022] The second learning unit 15 constructs a second estimation model using a deep learning technique that employs a neural network, using the common table data in the graph space processed by the preprocessing unit 12 as learning data. The learning method employed by the second learning unit 15 may be the same as that employed by the first learning unit 13. That is, the model learned by the second learning unit 15 has a feature vector and a variable vector for estimating the value of each item in the common table data.
[0023] The estimation unit 16 estimates missing data in the first table data and generates first table data in which the missing data has been complemented. In one example, the estimation unit 16 may complement the missing data in the first table data using a first estimation model constructed using the first estimation model. That is, the estimation unit 16 may acquire, from the first estimation model, a feature vector of a user and a variable vector of an item corresponding to the missing data in the first table data. The estimation unit 16 may estimate the missing data in the first table data based on the acquired feature vector and variable vector. Note that the common table data used by the second learning unit 15 described above may be generated based on the first table data in which the missing data has been complemented using the first estimation model.
[0024] The estimation unit 16 may complement missing data in the first table data using the second estimation model constructed by the second estimation model. That is, the estimation unit 16 may acquire a user feature vector and an item variable vector corresponding to the missing data in the first table data from the second estimation model. The estimation unit 16 may estimate the missing data in the first table data based on the acquired feature vector and variable vector. Note that the estimation unit 16 may complement missing data in the first table data using an average value between the estimated values estimated by the first estimation model and the estimated values estimated by the second estimation model.
[0025] Furthermore, the estimation unit 16 estimates missing data in the common table data and generates common table data in which the missing data is complemented. That is, the estimation unit 16 may acquire, from the second estimation model, a feature vector of a user and a variable vector of an item corresponding to the missing data in the common table data. The estimation unit 16 may estimate the missing data in the common table data based on the acquired feature vector and variable vector.
[0026] The estimation unit 16 estimates second data for a first user group other than the common users in the first table data. That is, in the illustrated example, the membership ranks, etc., of users with user IDs of 1, 5, 6, etc. are estimated. In one example, the estimation unit 16 may estimate the second data for the first user group other than the common users in the first table data using a first estimation model and a second estimation model. The estimation unit 16 may acquire feature vectors corresponding to each of the first user group other than the common users in the first table data from the first estimation model, and may acquire variable vectors of items corresponding to the second data from the second estimation model. The estimation unit 16 may estimate the second data for the first user group based on the acquired feature vectors and variable vectors.
[0027] The output unit 17 outputs integrated table data composed of the first data and the second data for the first user group. Missing data in the first data that constitutes the integrated data table may be complemented by the estimation unit 16. Missing data in the second data for the common users that constitute the integrated data table may be complemented by the estimation unit 16. The second data for the first user group that constitutes the integrated data table may be data estimated by the estimation unit 16. Note that the output unit 17 may output first table data in which the missing data has been complemented, or may output second table data in which the missing data has been complemented.
[0028] 6 is a flow diagram showing an example of the operation of the estimation device 10. First, in the estimation device 10, data constituting the first table data and data constituting the second table data are input to the input unit 11 (step S1). The second table data may include users other than the common user, but here it is assumed that the second table data is composed only of the common user.
[0029] Next, the first data constituting the first table data is preprocessed (step S2). That is, the preprocessing unit 12 represents the first table data in a graph space. Next, the first learning unit 13 constructs a first estimation model based on the preprocessed first table data (step S3). Next, the estimation unit 16 complements missing data in the first table data (step S4). That is, the estimation unit 16 complements missing data in the first table data using the feature vectors and variable vectors of the first estimation model constructed by the first learning unit 13.
[0030] Next, the common data is preprocessed (step S5). First, the preprocessing unit 12 extracts data of common users between the first table data and the second table data to generate common table data. Missing data in the first table data is complemented in step S4. Next, the preprocessing unit 12 represents the generated common table data in a graph space. Next, the second learning unit 15 constructs a second estimation model based on the preprocessed common table data (step S6).
[0031] Next, the estimation unit 16 complements the missing data in the second table data (step S7). That is, the estimation unit 16 complements the missing data in the second table data by using the feature vectors and variable vectors of the second estimation model constructed by the second learning unit 15.
[0032] Next, the estimation unit 16 re-complements the missing data in the first table data (step S8). For example, the estimation unit 16 estimates the missing data in the first table data using the feature vectors of the first estimation model and the variable vectors of the second estimation model, and complements the missing data by averaging the estimated missing data and the missing data estimated in step S4. Note that the first data for the common user in the first table data may be complemented using the feature vectors and variable vectors of the second estimation model constructed by the second learning unit 15.
[0033] Next, the estimation unit 16 estimates the second data of the first user group (step S9). That is, the estimation unit 16 estimates the second data of the first user group excluding the common user by using the feature vector of the first estimation model and the variable vector of the second estimation model.
[0034] Subsequently, the output unit 17 outputs the integrated table data in which the missing data and the second data for the first user group have been complemented by the estimation unit 16. In one example, the missing data in the first table data portion has been complemented in step S8, the missing data in the second table data portion has been complemented in step S7, and the second data for the first user group excluding the common user has been estimated in step S9.
[0035] As described above, the example estimation device 10 includes a first learning unit 13 that constructs a first estimation model that estimates missing data in the first table data based on the first table data, a second learning unit 15 that constructs a second estimation model that estimates missing data of the common user group in the second table data based on second data about the common user group in the second table data and first data about the common user group in the first table data, and an estimation unit 16 that estimates the value of the second data about the first user group excluding the common user group based on the first estimation model and the second estimation model.
[0036] In the above-described estimation device 10, missing data in the first table data is complemented by the first estimation model, and missing data in the second table data is complemented by the second estimation model. Then, second data for a first user other than a common user is complemented based on the first estimation model and the second estimation model. The second estimation model that estimates the missing data in the second table data is constructed based on the first data and the second data and reflects the contents of the first data. Therefore, the estimation unit 16 that estimates the missing data based on the first estimation model and the second estimation model can accurately estimate second data that is not originally included in the first table data.
[0037] An example first estimation model may estimate the value of each item corresponding to each user in the first user group from the feature quantities of each user in the first user group and the feature quantities of each item constituting the first table data. An example second estimation model may estimate the value of each item corresponding to each user in the common user group from the feature quantities of each user in the common user group and the feature quantities of each item constituting the second table data. An example estimation unit 16 may estimate the value of the second data for the first user group excluding the common user group based on the feature quantities of each user in the first user group used in the first estimation model and the feature quantities of each item constituting the second table data used in the second estimation model. In this configuration, when estimating the second data for the first user group, the user feature quantities reflect the data of the first table data, and the feature quantities of the items of the second data reflect the data of the common users in the first table data and the second table data. In this case, since the user feature quantities are generated using the first table data composed of the first user group, which is larger than the common user group, bias in the user characteristics is less likely to occur. Furthermore, since the feature quantities of the items of the second data are generated based on the first data and the second data of the common user group, the feature quantities can be generated more accurately than when the feature quantities of the items of the second data are generated based only on the second data of the common user group. Therefore, the second data that is not originally included in the first table data can be estimated with high accuracy based on the feature quantities of the first user and the feature quantities of the items of the second data.
[0038] In one example, the feature values of each user in the first user group used in the first estimation model and the feature values of each item constituting the second table data used in the second estimation model may be expressed as vectors. This configuration makes it easy to build an estimation model using so-called graph neural network technology.
[0039] In one example, the second estimation model may estimate missing data in the first table data based on the first data and the second data for a common user group. In this configuration, by increasing the number of items that are data strings, the missing data can be estimated more accurately than with the first estimation model.
[0040] In one example, the second learning unit 15 may construct the second estimation model based on the first table data in which missing data has been complemented by the first estimation model, and the second table data. In this configuration, by using the first table data in which missing data has been complemented, it is possible to construct the second estimation model with high estimation accuracy.
[0041] Although one embodiment has been described above in detail with reference to the drawings, the specific configuration is not limited to this embodiment. For example, although an example of constructing a learning model using deep learning technology with a graph neural network has been described, other machine learning technology may be used as long as it can estimate the second data about the first user based on the features of the first user estimated from the first estimation model and the features of the second data item estimated from the second estimation model.
[0042] The block diagrams used to explain the above embodiments show functional blocks. These functional blocks (components) are realized by any combination of hardware and / or software. Furthermore, the method for realizing each functional block is not particularly limited. That is, each functional block may be realized using a single device that is physically or logically coupled, or may be realized using two or more physically or logically separated devices that are connected directly or indirectly (e.g., via wire, wirelessly, etc.) and these multiple devices. The functional block may also be realized by combining the single device or multiple devices with software.
[0043] Functions include, but are not limited to, judgment, determination, assessment, 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, allocating, mapping, and assignment. For example, a functional block (component) that performs transmission is called a transmitting unit or transmitter. As mentioned above, there are no particular limitations on how these functions are implemented.
[0044] For example, the estimation device 10 according to an embodiment of the present disclosure may function as a computer that performs the estimation processing of the present disclosure. Fig. 7 is a diagram illustrating an example of a hardware configuration of the estimation device 10 according to an embodiment of the present disclosure. The above-described 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, etc.
[0045] In the following description, the term "apparatus" can be interpreted as a circuit, a device, a unit, etc. The hardware configuration of the input unit 11, the preprocessing unit 12, the first learning unit 13, the second learning unit 15, the estimation unit 16, and the output unit 17 may be configured to include one or more of the devices shown in the figure, or may be configured to exclude some of the devices.
[0046] Each function in the estimation device is realized by loading specified software (programs) onto hardware such as processor 1001 and memory 1002, causing processor 1001 to perform calculations, control communication via communication device 1004, and control at least one of reading and writing data in memory 1002 and storage 1003.
[0047] The processor 1001 controls the entire computer by running, for example, an operating system. The processor 1001 may be configured by a central processing unit (CPU) including an interface with peripheral devices, a control device, an arithmetic unit, a register, etc. For example, the preprocessing unit 12, the first learning unit 13, the second learning unit 15, the estimation unit 16, etc. may be realized by the processor 1001.
[0048] The processor 1001 also reads programs (program codes), software modules, data, etc. from at least one of the storage 1003 and the communication device 1004 into the memory 1002 and executes various processes in accordance with these programs. The programs used are those that cause a computer to execute at least some of the operations described in the above-described embodiments. For example, the preprocessing unit 12 may be implemented by a control program stored in the memory 1002 and running on the processor 1001, and similar implementations may be made for other functional blocks. While the above-described various processes have been described as being executed by one processor 1001, they may also 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 also be transmitted from a network via a telecommunications line.
[0049] The memory 1002 is a computer-readable recording medium and may be configured by, for example, at least one of 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 also be called a register, a cache, a main memory (primary storage device), etc. The memory 1002 can store executable programs (program codes), software modules, etc. for implementing an estimation method according to an embodiment of the present disclosure.
[0050] 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 (registered trademark) disk), a smart card, a flash memory (e.g., a card, a stick, a key drive), a floppy (registered trademark) disk, a magnetic strip, etc. Storage 1003 may also be referred to as an auxiliary storage device. The above-mentioned storage medium may be, for example, a database, a server, or other appropriate medium including at least one of memory 1002 and storage 1003.
[0051] The communication device 1004 is hardware (transmission / reception device) for communicating between computers via at least one of a wired network and a wireless network, and is also referred to as, for example, a network device, a network controller, a network card, or a communication module. The communication device 1004 may be configured to include a high-frequency switch, a duplexer, a filter, a frequency synthesizer, etc. to realize at least one of frequency division duplex (FDD) and time division duplex (TDD). For example, the input unit 11, the output unit 17, etc. described above may be realized by the communication device 1004.
[0052] The input device 1005 is an input device (e.g., a keyboard, a mouse, a microphone, a switch, a button, a sensor, etc.) that receives input from the outside. The output device 1006 is an output device (e.g., a display, a speaker, an LED lamp, etc.) that outputs to the outside. Note that the input device 1005 and the output device 1006 may be integrated into one device (e.g., a touch panel).
[0053] 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 using a single bus, or may be configured using different buses between each device.
[0054] The estimation device may also 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 using at least one of these pieces of hardware.
[0055] The notification of information is not limited to the aspects / embodiments described in the present disclosure and may be performed using other methods. For example, the notification of information may be performed by physical layer signaling (e.g., Downlink Control Information (DCI) and Uplink Control Information (UCI)), higher layer signaling (e.g., Radio Resource Control (RRC) signaling, Medium Access Control (MAC) signaling, broadcast information (Master Information Block (MIB) and System Information Block (SIB))), other signals, or a combination thereof. Furthermore, the RRC signaling may be referred to as an RRC message, and may be, for example, an RRC Connection Setup message, an RRC Connection Reconfiguration message, or the like.
[0056] The order of the procedures, sequences, flowcharts, etc. of each aspect / embodiment described in this disclosure may be changed unless it is consistent. For example, the methods described in this disclosure present elements of various steps using an example order, and are not limited to the particular order presented.
[0057] Input and output information may be stored in a specific location (for example, memory) or may be managed using a management table. Input and output information may be overwritten, updated, or added to. Output information may be deleted. Input information may be sent to another device.
[0058] The determination may be made based on a value represented by one bit (0 or 1), a Boolean value (true or false), or a numerical comparison (e.g., comparison with a predetermined value).
[0059] The aspects / embodiments described in this disclosure may be used alone, in combination, or switched depending on the implementation. Notification of predetermined information (e.g., notification that "X is true") is not limited to explicit notification, but may be implicit (e.g., not notifying the predetermined information).
[0060] 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 herein. 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.
[0061] 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.
[0062] Software, instructions, information, etc. may also be transmitted or received over a transmission medium. For example, if software is transmitted from a website, server, or other remote source using wired technologies (such as coaxial cable, fiber optic cable, twisted pair, Digital Subscriber Line (DSL)), and / or wireless technologies (such as infrared, microwave), then these wired and / or wireless technologies are included within the definition of transmission media.
[0063] The information, signals, etc. described in this disclosure may be represented using any of a variety of different technologies. For example, 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.
[0064] In addition, terms explained in this disclosure and terms necessary for understanding this disclosure may be replaced with terms having the same or similar meanings.
[0065] Furthermore, the information, parameters, etc. described in the present disclosure may be expressed using absolute values, may be expressed using relative values from a predetermined value, or may be expressed using other corresponding information. For example, a radio resource may be indicated by an index.
[0066] 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, searching, inquiring (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 memory), and the like. Furthermore, "judgment" and "decision" can include regarding resolving, selecting, choosing, establishing, comparing, etc. as having been "judged" or "decided." In other words, "judgment" and "decision" can include regarding some action as having been "judged" or "decided." Furthermore, "judgment (decision)" can be interpreted as "assuming," "expecting," "considering," etc.
[0067] The terms "connected," "coupled," or any variation thereof, refer to any direct or indirect connection or coupling between two or more elements, and may include the presence of one or more intermediate elements between two elements that are "connected" or "coupled" to each other. The coupling or connection between elements may be physical, logical, or a combination thereof. For example, "connected" may be read as "access." As used in this disclosure, two elements may be considered to be "connected" or "coupled" to each other using one or more wires, cables, and / or printed electrical connections, as well as electromagnetic energy having wavelengths in the radio frequency range, microwave range, and optical (both visible and invisible) range, as some non-limiting and non-exhaustive examples.
[0068] 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."
[0069] As used in this disclosure, any reference to an element using a designation such as "first," "second," etc. does not generally limit the quantity or order of those elements. These designations may be used in this disclosure 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 or that the first element must in some way precede the second element.
[0070] When the terms "include," "including," and variations thereof are used in this disclosure, these terms are intended to be inclusive, similar to the term "comprising." Furthermore, when the term "or" is used in this disclosure, it is not intended to be an exclusive or.
[0071] In this disclosure, where articles are added by translation, such as a, an, and the in English, the disclosure may include that the nouns following these articles are in the plural form.
[0072] In the present disclosure, the term "A and B are different" may mean "A and B are different from each other." The term may also mean "A and B are each different from C." Terms such as "separate" and "coupled" may also be interpreted in the same way as "different."
[0073] An estimation device according to the present disclosure has the following configuration: [1] An estimation device including: a first learning unit that constructs a first estimation model that estimates missing data in first table data based on first table data configured with first data indicating attributes of each user in the first user group, a second learning unit that constructs a second estimation model that estimates missing data of a common user group in the second table data based on second table data having second data indicating attributes of each user in a second user group that is a common user group common to a part of the first user group, and the first data for the common user group in the first table data, and an estimation unit that estimates a value of the second data for the first user group excluding the common user group based on the first estimation model and the second estimation model. [2] The estimation device described in [1], wherein the first estimation model estimates a value of each item corresponding to each user of the first user group from a feature amount of each user of the first user group and a feature amount of each item constituting the first table data, the second estimation model estimates a value of each item corresponding to each user of the common user group from a feature amount of each user of the common user group and a feature amount of each item constituting the second table data, and the estimation unit estimates a value of the second data for the first user group excluding the common user group based on the feature amount of each user of the first user group used in the first estimation model and the feature amount of each item constituting the second table data used in the second estimation model. [3] The estimation device described in [2], wherein the feature amount of each user of the first user group used in the first estimation model and the feature amount of each item constituting the second table data used in the second estimation model are expressed as vectors. [4] The estimation device according to any one of [1] to [3], wherein the second estimation model estimates missing data in the first table data based on the first data and the second data for the common user group.[5] The estimation device according to any one of [1] to [4], wherein the second learning unit constructs the second estimation model based on the first table data in which the missing data has been complemented by the first estimation model and the second table data.
[0074] 10...estimation device, 11...input unit, 11A...first data input unit, 11B...second data input unit, 12...preprocessing unit, 13...first learning unit, 15...second learning unit, 16...estimation unit, 17...output unit, 1001...processor, 1002...memory, 1003...storage, 1004...communication device, 1005...input device, 1006...output device, 1007...bus.
Claims
1. An estimation device comprising: a first learning unit that constructs a first estimation model that estimates missing data in the first table data based on first table data constituted by first data indicating attributes of each user in a first user group; a second learning unit that constructs a second estimation model that estimates missing data of the common user group in the second table data based on second table data having second data indicating attributes of each user of a second user group, which is a common user group common to a part of the first user group, and the first data for the common user group in the first table data; and an estimation unit that estimates a value of the second data for the first user group excluding the common user group based on the first estimation model and the second estimation model.
2. The estimation device described in claim 1, wherein the first estimation model estimates values of each item corresponding to each user of the first user group from features of each user of the first user group and features of each item constituting the first table data, the second estimation model estimates values of each item corresponding to each user of the common user group from features of each user of the common user group and features of each item constituting the second table data, and the estimation unit estimates values of the second data for the first user group excluding the common user group based on the features of each user of the first user group used in the first estimation model and the features of each item constituting the second table data used in the second estimation model.
3. The estimation device described in claim 2, wherein the features of each user of the first user group used in the first estimation model and the features of each item constituting the second table data used in the second estimation model are expressed as vectors.
4. The estimation device according to claim 1, wherein the second estimation model estimates missing data in the first table data based on the first data and the second data for the common user group.
5. The estimation device according to claim 1, wherein the second learning unit constructs the second estimation model based on the first table data in which the missing data has been complemented by the first estimation model and the second table data.
6. An estimation device comprising: a first learning unit that constructs a first estimation model that estimates missing data in the first table data based on first table data constituted by first data indicating attributes of each user in the first user group; and a second learning unit that constructs a second estimation model that estimates missing data of the common user group in the second table data based on second table data having second data indicating attributes of each user in a second user group, which is a common user group common to a part of the first user group, and the first data for the common user group in the first table data in which the missing data has been complemented by the first estimation model.
Citation Information
Patent Citations
Apparatus, method and program for integrating databases, and data complementer
JP2020140493A
Pipeline accident risk evaluation device and pipeline accident risk evaluation method
JP2022118672A
Methods and apparatus for transforming and statistically modeling relational databases to synthesize privacy-protected anonymized data
US20180165475A1
