Device and method

Abstract

A device according to an exemplary embodiment is provided with: a model generation unit that, on the basis of performance data and each of a plurality of sets of extension data for extending the performance data to a period preceding the data acquisition period for the performance data, generates a plurality of future prediction models that predict fluctuations in performance data after the data acquisition period; and a model selection unit that selects, from among the plurality of future prediction models, a future prediction model having relatively high prediction accuracy.

Description

Apparatus and method

[0001] The present disclosure relates to an apparatus and method for predicting future data.

[0002] Patent Literature 1 discloses a technique for predicting future data after the actual data by using a prediction model generated based on time-series actual data.

[0003] International Publication No. 2019-065610

[0004] There is a known technology for making future predictions using a prediction model generated based on time-series performance data. However, when such technology is used, it is difficult to generate a highly accurate prediction model if the amount of performance data is insufficient, i.e., if the period of time during which performance data is accumulated is short. Therefore, a technology is needed that can accurately predict future data even when the amount of performance data is small.

[0005] An object of the present disclosure is to provide an apparatus and method capable of predicting future data with high accuracy.

[0006] An apparatus according to one aspect of the present disclosure includes a data extension unit that acquires multiple pieces of extended data for extending actual data to a period prior to the data acquisition period of the actual data; a model generation unit that generates multiple future prediction models that predict fluctuations in the actual data after the data acquisition period based on each of the multiple pieces of extended data and the actual data; a model selection unit that selects a future prediction model with relatively high prediction accuracy from the multiple future prediction models; and a prediction data generation unit that predicts actual data after the data acquisition period using the extended data used to generate the future prediction model selected by the model selection unit.

[0007] In the above device, not only actual data but also extended data, which is actual data extended into the past, is used to generate a future prediction model. In this way, data from a longer period is used to generate a learning model, making it easier for information such as seasonality to be reflected in the learning model, thereby improving prediction accuracy. Furthermore, the above device selects a model with high prediction accuracy from multiple future prediction models. This makes it possible to eliminate future prediction models generated based on extended data that do not match the trends of the actual data, thereby improving prediction accuracy. Therefore, the above device can predict future data with high accuracy.

[0008] According to one aspect of the present disclosure, it is possible to provide an apparatus and a method capable of predicting future data with high accuracy.

[0009] FIG. 1 is a schematic diagram showing the configuration of an example prediction device. FIG. 2 is a diagram for explaining an example of performance data. FIG. 3 is a diagram for explaining an example of a state in which performance data has been expanded into the past. FIG. 4 is a diagram for explaining an example of future prediction data. FIG. 5 is a flow chart for explaining the processing of an example prediction device. FIG. 6 is a diagram showing an example of a device constituting the prediction device.

[0010] Hereinafter, exemplary embodiments will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements are designated by the same reference numerals, and redundant description will be omitted.

[0011] FIG. 1 is a schematic diagram showing the configuration of an example prediction device. The prediction device 10 is a device that predicts future data following time-series performance data. In one exemplary embodiment, a prediction device that predicts the future number of subscribers who will subscribe to a first service will be described. The service may be the provision of tangible or intangible value to subscribers. As shown in FIG. 1 , the example prediction device 10 includes a performance data acquisition unit 11, a data expansion unit 12, a model generation unit 13, a model selection unit 15, a prediction data generation unit 16, and a prediction data output unit 17.

[0012] The performance data acquisition unit 11 acquires performance data to be predicted. FIG. 2 is a diagram illustrating an example of performance data. The performance data D0 may be, for example, time-series data accumulated over a certain period (data acquisition period T0). The performance data D0 may be accumulated in another server device, database, or the like, and may be transmitted from the server device or the like to the prediction device 10. The data acquisition period T0 for the performance data D0 may be divided into a first period T1 and a second period T2. The performance data D0 includes prediction performance data D1, which is data from the first period T1, and verification performance data D2, which is data from the second period T2 that follows the first period T1. The prediction performance data D1 and the verification performance data D2 may be temporally continuous data. For example, the second period T2 corresponding to the verification performance data D2 may be approximately two weeks to one month. In the illustrated example, the data acquisition period T0 is three months, the first period T1 is two months, and the second period T2 is one month. The lengths of the first period T1 and the second period T2 are not particularly limited.

[0013] If the performance data D0 is the number of subscribers to the first service, the performance data D0 may be composed of a quantitative variable indicating new subscribers for each period. For example, if the first service has multiple service plans, the performance data D0 may be composed of the actual number of subscribers for each service plan during a unit period, aggregated by prefecture. In this case, the performance data D0 may include prefecture information and service plan information as qualitative variables. The performance data D0 may also be the total number of actual subscribers during a unit period. Note that in the performance data D0 shown in FIG. 2, the total number of subscribers nationwide for each day is plotted as a value.

[0014] The data extension unit 12 acquires multiple pieces of extension data for extending the historical data D0 to a period (data extension period T3) prior to (past) the data acquisition period T0 of the historical data D0. FIG. 3 is a diagram illustrating an example of a state in which historical data has been extended to the past. In one exemplary embodiment, for convenience, the data extension unit 12 acquires two pieces of extension data (first extension data and second extension data). Note that while FIG. 3 shows only one piece of extension data D3 (e.g., the first extension data), the data extension unit 12 also acquires other pieces of extension data (e.g., the second extension data). The extension data D3 may be generated by referencing similar data having a similar tendency to the historical data D0. The similar data is time-series data for the data acquisition period T0 and the data extension period T3.

[0015] The length of the data extension period T3 is not particularly limited. In one example, the start of the data extension period T3 may be a point going back one year or more from the end of the first period T1 of the actual data D0. That is, if the first period T1 is from April 1 to May 31 of the current year, for example, the start of the data extension period T3 may be before June 1 of the previous year. In this case, the similar data may include time-series data from June 1 of the previous year to June 30 of the current year.

[0016] For example, if a certain correlation is observed between the trends of the transitions of two pieces of time-series data, the two pieces of time-series data have a tendency to be similar to each other. In other words, having a similar tendency may mean that there is a positive correlation between the actual data D0 and the data of the similar data during the data acquisition period T0. Also, having a similar tendency may mean that when the actual data D0 and the data of the similar data during the data acquisition period T0 are normalized and plotted to visualize them, the image features of the actual data D0 and the similar data are similar. Note that known techniques may be used to determine the similarity of image features.

[0017] 2, the number of subscribers increases in the first period T1 and decreases in the second period T2. Therefore, time-series data in which the number of subscribers increases in the first period T1 and decreases in the second period T2 can be said to be similar data having a similar trend to the performance data D0.

[0018] In one example, the similar data may be time-series data showing the transition of the number of subscribers of a second service different from the first service. If the second service has multiple service plans, the similar data may be configured by the number of subscribers for each service plan during a unit period, aggregated by prefecture. In this case, the similar data may include prefecture information and service plan information as qualitative variables.

[0019] The first extended data may be generated by referring to the fluctuation ratio of the similar data in the data extension period T3. In one example, the ratio (actual ratio) of the total number of subscribers of the prediction actual data D1 to the total number of subscribers of the similar data in the first period T1 is calculated. The first extended data is derived by multiplying the similar data in the data extension period T3 by the actual ratio. This generates first extended data having a fluctuation ratio that matches the fluctuation ratio of the similar data. The total number of prediction actual data may be calculated, for example, for each service plan. In this case, extended data is derived for each service plan.

[0020] For example, consider a case where the historical data D0 is time-series data on the number of subscribers of the first service plan and the second service plan constituting the first service. If the total number of subscribers of the first service plan in the first period T1 is X, the total number of subscribers of the second service plan in the first period T1 is Y, and the total number of subscribers in the similar data in the first period T1 is Z, then the historical ratio of the first service plan to the similar data in the first period T1 is (X / Z), and the historical ratio of the second service plan to the similar data in the first period T1 is (Y / Z). In this case, the extended data of the first service plan is derived by multiplying the similar data in the data extension period T3 by (X / Z), and the extended data of the second service plan is derived by multiplying the similar data in the data extension period T3 by (Y / Z). In addition, when the similar data for the data extension period T3 is multiplied by ((X+Y) / Z), extension data for the entire performance data obtained by adding together the first service plan and the second service plan is derived.

[0021] Furthermore, the second extended data may be generated using a past prediction model generated based on the similar data and the actual data D0. The past prediction model predicts extended data corresponding to data in a data extension period T3 of the similar data by learning the relationship between the similar data in the first period T1 and the prediction actual data D1. For example, the past prediction model may be a regression model generated using quantitative and qualitative variables in the similar data as explanatory variables and the quantitative variable in the prediction actual data D1 as a target variable. Note that, for example, a past prediction model that predicts the number of subscribers for each service plan may be generated by including service plan information of the actual data in the explanatory variables.

[0022] As described above, the first extended data and the second extended data may be generated by referring to one similar data. Alternatively, the first extended data and the second extended data may be generated by referring to different similar data. In this case, the first extended data and the second extended data may be generated by referring to a fluctuation ratio of the similar data or may be generated based on a past prediction model.

[0023] The model generator 13 generates a plurality of future prediction models that predict future fluctuations in the performance data based on each of the plurality of extended data and the performance data. The plurality of future prediction models may be time series prediction models generated by a learning method such as Prophet, ARIMA (Auto Regressive Integrated Moving Average), or MAML (Model-Agnostic Meta-Learning). In one exemplary embodiment, the model generator 13 generates a first future prediction model based on the first extended data and the performance data, and generates a second future prediction model based on the second extended data and the performance data.

[0024] In one example, the model generation unit 13 generates a first future prediction model that predicts the number of subscribers after the first period T1 using the first extended data and the prediction actual data D1 as learning data. The model generation unit 13 also generates a second future prediction model that predicts the number of subscribers after the first period T1 using the second extended data and the prediction actual data D1 as learning data. The first future prediction model and the second future prediction model may be generated using different learning methods or the same learning method.

[0025] The model generation unit 13 can generate a future prediction model that predicts data trends in a period after the data acquisition period T0, using the extended data D3 and the performance data D0 as learning data.

[0026] The model selection unit 15 selects a future prediction model with relatively high prediction accuracy from among multiple future prediction models. For example, the model selection unit 15 selects a future prediction model with relatively high prediction accuracy based on a comparison between prediction results for the second period T2 derived based on the multiple future prediction models and the verification actual data D2. In one exemplary embodiment, the model selection unit 15 derives error rates for the prediction results based on the first future prediction model and the prediction results based on the second future prediction model with respect to the verification actual data D2. The model selection unit 15 determines the future prediction model with the smaller error rate as the model to be used for generating prediction data.

[0027] The method for deriving the error rate is not particularly limited. In one example, the error rate E may be derived by the following formula, where P is the total number of subscribers in the prediction result and T is the total number of subscribers in the verification performance data D2: E=(P−T) / T

[0028] The prediction data generation unit 16 performs data prediction for a prediction period T4 after the data acquisition period T0 based on the future prediction model selected by the model selection unit 15. FIG. 4 is a diagram for explaining an example of future prediction data. The predicted future data D4 is the number of subscribers who will subscribe to the first service during the prediction period T4 after the data acquisition period T0. The number of subscribers may be derived for each service plan. Alternatively, the number of subscribers may be the cumulative number of subscribers.

[0029] When three or more future prediction models are generated in the model generation unit 13, the model selection unit 15 may select the future prediction models in descending order of prediction accuracy. In this case, the prediction data generation unit 16 may derive the average value of the prediction data predicted based on the multiple future prediction models selected by the model selection unit 15 as the prediction result.

[0030] The forecast data output unit 17 outputs the forecast values ​​derived by the forecast data generation unit 16. For example, the forecast data output unit 17 may output the forecast values ​​as a graph with the horizontal axis representing time and the vertical axis representing the number of subscribers. In this case, the actual data and the forecast data may be depicted on a single graph in a manner that allows them to be distinguished by color or the like.

[0031] 5 is a flowchart illustrating the process of the exemplary prediction device 10. In the process of the exemplary prediction device 10, first, performance data D0 is acquired (step S1). For example, when multiple types of performance data are stored in a server device or the like, a user may specify (input) data to be predicted, and the specified performance data D0 may be acquired from the server device or the like by the prediction device 10.

[0032] Next, the prediction device 10 extends the acquired performance data D0 to a time period prior to (past) the data acquisition period T0 (step S2). That is, the data is extended to a period prior to the start of service provision, which does not actually exist. In step S2, similar data having a similar tendency to the performance data D0 is acquired, and multiple extended data are acquired based on the similar data. For example, if a database is prepared in which performance data for multiple different services is accumulated, data similar to the performance data D0 may be acquired as the similar data from among the multiple performance data having a start time earlier than the performance data D0 to be predicted.

[0033] Next, a future prediction model is generated (step S3). The future prediction model is a model that predicts trends in the actual data after the data acquisition period T0 based on the extended data and the actual data. For example, if the first extended data and the second extended data have been acquired in step S2, a first future prediction model is generated based on the forecast actual data D1 and the first extended data, and a second future prediction model is generated based on the forecast actual data D1 and the second extended data.

[0034] Next, a future prediction model is selected (step S4). In step S4, the prediction results for the second period T2 predicted by each future prediction model are compared with the verification performance data D2, and a model with a low error rate is selected.

[0035] Next, prediction data is generated (step S5). That is, the future prediction model selected in step S4 predicts the trend in subscriber numbers for a period after the data acquisition period T0. Then, the prediction data is output (step S6). The data output may be, for example, a graph plotting the number of subscribers for each unit period, or a graph showing the cumulative number of subscribers. Text data such as the number of subscribers may also be output. Note that in step S5, a new future prediction model for predicting data after the data acquisition period T0 may be generated based on the extended data and actual data D0 used to generate the future prediction model selected in step S4.

[0036] As described above, the prediction device 10 according to one aspect of the present disclosure includes a data extension unit 12, a model generation unit 13, a model selection unit 15, and a prediction data generation unit 16. The data extension unit 12 acquires multiple pieces of extension data D3 for extending actual data D0 to a period prior to the data acquisition period T0. The model generation unit 13 generates multiple future prediction models for predicting fluctuations in the actual data after the data acquisition period T0 based on each of the multiple pieces of extension data D3 and the actual data. The model selection unit 15 selects a future prediction model with relatively high prediction accuracy from the multiple future prediction models. The prediction data generation unit 16 predicts actual data after the data acquisition period T0 using the extension data used to generate the future prediction model selected by the model selection unit 15. Note that the prediction data generation unit 16 may use the future prediction model selected by the model selection unit 15 as is, or may generate a new future prediction model using the extension data of the future prediction model selected by the model selection unit 15.

[0037] Generally, techniques are known for predicting future trends in time-series performance data using a prediction model generated based on the performance data. However, for example, in the early stages of a new service launch, there is insufficient accumulated performance data for the service, making it difficult to generate a prediction model with high prediction accuracy. In a prediction device 10 according to one exemplary embodiment, not only performance data but also extended data D3, which extends the performance data into the past, are used to generate a future prediction model. In this way, using data from a longer period to generate a learning model makes it easier to reflect information such as seasonality in the learning model, thereby improving prediction accuracy. In particular, when the length from the start of the extended data to the end of the forecast performance data is one year or more, trends in data trends throughout the year are more likely to be reflected. Furthermore, the prediction device 10 selects a model with high prediction accuracy from multiple future prediction models based on a comparison with verification performance data D2, which is part of the performance data D0. This allows future prediction models generated based on extended data that do not match the trends in the performance data to be eliminated, thereby improving prediction accuracy. Therefore, the prediction device 10 can accurately predict future data.

[0038] The data extension unit 12 may acquire the extended data D3 generated by referring to similar data that has a similar tendency to the performance data D0. In this configuration, the extended data D3 is generated based on the similar data, so that the extended data D3 that is likely to match the trend of the performance data can be efficiently generated.

[0039] The historical data D0 may include prediction historical data D1, which is data for a first time period T1, and verification historical data D2, which is data for a second time period T2 that follows the first time period T1. The model generator 13 may generate multiple future prediction models based on each of the multiple pieces of extended data D3 and the prediction historical data D1. In this configuration, generating multiple future prediction models can increase the probability of obtaining a model with high prediction accuracy.

[0040] The model selection unit 15 may select a future prediction model with relatively high prediction accuracy based on a comparison between prediction results for the second period T2 derived from each of the multiple future prediction models and the verification actual data D2. In this configuration, each future prediction model is compared with the verification actual data D2, which is part of the actual data D0, so that a future prediction model with high prediction accuracy can be selected with high reliability.

[0041] There may be a positive correlation between the performance data D0 and the similar data. In this configuration, similar data that is similar to the performance data D0 can be objectively acquired.

[0042] The data expansion unit 12 may acquire multiple pieces of expanded data D3 by respectively referencing multiple pieces of similar data that are different from each other. In an exemplary embodiment, the prediction device 10 determines whether actual data and similar data are similar based on the similarity trend with the actual data D0 during the data acquisition period T0. In this case, there is no guarantee that the trend of the similar data during the data expansion period T3 is suitable for expanding the actual data. Generating multiple pieces of expanded data based on multiple pieces of similar data can generate expanded data with more diverse trends than generating multiple pieces of expanded data based on a single piece of similar data. This can lead to the generation of a more accurate future prediction model. The above configuration makes it easy to acquire multiple pieces of expanded data with different fluctuation trends during the data expansion period.

[0043] The fluctuation ratio of at least one of the plurality of extended data D3 may match the fluctuation ratio of the similar data. In this configuration, the trend of the extended data D3 can be matched with the trend of the similar data.

[0044] At least one of the plurality of extended data D3 may be generated using a past prediction model generated based on similar data and performance data. In this configuration, the similarity between the similar data and performance data in the data acquisition period T0 can be reflected in the extended data D3.

[0045] In one exemplary embodiment, a prediction device 10 that estimates the transition in the number of subscribers to a service has been described. Here, the service may be, for example, the provision of social infrastructure such as electricity, gas, or communications, but the type of service is not particularly limited. Furthermore, the prediction target of the prediction device is not limited to the transition in the number of subscribers to the service. The prediction target of the prediction device may be any time-series data, such as the number of sales or sales of goods, etc.

[0046] The block diagrams used to explain the above embodiments show functional blocks. These functional blocks (components) are realized by any combination of at least one of hardware and 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 directly or indirectly connected (e.g., wired, wireless, etc.) and these multiple devices. The functional block may also be realized by combining software with the single device or multiple devices.

[0047] 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.

[0048] For example, the prediction device 10 according to an embodiment of the present disclosure may function as a computer that performs the processing of the present disclosure. Fig. 6 is a diagram illustrating an example of a hardware configuration of the prediction device 10 according to an embodiment of the present disclosure. The prediction device 10 described above 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.

[0049] In the description, the term "apparatus" can be interpreted as a circuit, a device, a unit, etc. The hardware configuration of the prediction apparatus 10 may be configured to include one or more of the apparatuses shown in the drawings, or may be configured to exclude some of the apparatuses.

[0050] Each function of the prediction device 10 is realized by loading specific software (programs) onto hardware such as a processor 1001 and a memory 1002, causing the processor 1001 to perform calculations, control communication via a communication device 1004, and control at least one of reading and writing data in the memory 1002 and the storage 1003.

[0051] 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, the above-mentioned data expansion unit 12 and the like may be realized by the processor 1001.

[0052] The processor 1001 also reads programs (program code), 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. The program used is a program that causes a computer to execute at least some of the operations described in the above-described embodiments. For example, the data expansion unit 12 may be implemented by a control program stored in the memory 1002 and running on the processor 1001, and similar implementations may also 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.

[0053] 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 read-only memory (EPROM), an electrically erasable programmable read-only memory (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 a wireless communication method according to an embodiment of the present disclosure.

[0054] 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.

[0055] 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, a communication module, etc. 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).

[0056] 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).

[0057] 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.

[0058] The prediction device 10 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.

[0059] Notification of information is not limited to the aspects / embodiments described in this disclosure, and may be performed using other methods.

[0060] 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.

[0061] 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 transmitted to another device.

[0062] 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).

[0063] 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).

[0064] 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.

[0065] 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.

[0066] 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.

[0067] 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.

[0068] 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.

[0069] As used in this disclosure, the terms "system" and "network" are used interchangeably.

[0070] Furthermore, the information, parameters, etc. described in this 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.

[0071] The names used for the above parameters are not limiting in any way, and furthermore, the mathematical formulas etc. using these parameters may differ from those explicitly disclosed in this disclosure.

[0072] 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 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, all of which are considered to be "determining." "Determining" and "determining" may also include resolving, selecting, choosing, establishing, comparing, and the like, all of which are considered to be "determining." In other words, "judgment" and "decision" can include regarding some action as having been "judged" or "decided." Also, "judgment (decision)" can be interpreted as "assuming," "expecting," "considering," etc.

[0073] 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.

[0074] 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."

[0075] 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.

[0076] The "unit" in the configuration of each of the above devices may be replaced with "means," "circuit," "device," etc.

[0077] 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.

[0078] 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.

[0079] 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."

[0080] 10... prediction device (device), 12... data extension unit, 13... model generation unit, 15... model selection unit, 16... prediction data generation unit.

Claims

1. An apparatus comprising: a data extension unit that acquires a plurality of extension data for extending actual data to a period prior to a data acquisition period of the actual data; a model generation unit that generates a plurality of future prediction models that predict fluctuations in the actual data after the data acquisition period based on each of the plurality of extension data and the actual data; a model selection unit that selects a future prediction model with relatively high prediction accuracy from the plurality of future prediction models; and a prediction data generation unit that predicts the actual data after the data acquisition period using the extension data used to generate the future prediction model selected by the model selection unit.

2. The device according to claim 1, wherein the data expansion unit acquires the expanded data generated by referring to similar data that has a tendency to be similar to the performance data.

3. The device described in claim 1, wherein the performance data includes performance data for prediction, which is data for a first period, and performance data for verification, which is data for a second period after the first period, and the model generation unit generates the multiple future prediction models based on each of the multiple extended data and the performance data for prediction.

4. The device described in claim 3, wherein the model selection unit selects the future forecasting model with relatively high forecasting accuracy based on a comparison between the forecast results for the second period derived based on the multiple future forecasting models and the verification actual data.

5. The device according to claim 2, wherein there is a positive correlation between the performance data and the similar data.

6. The device according to claim 2, wherein the data extension unit acquires the plurality of pieces of extended data by referring to the plurality of similar data that are different from each other.

7. The device of claim 2, wherein the variation ratio of at least one of the plurality of extended data matches the variation ratio of the similar data.

8. The device according to claim 2, wherein at least one of the plurality of extended data is generated using a past prediction model generated based on the similar data and the performance data.

9. A method comprising the steps of: acquiring a plurality of extension data for extending actual data to a period prior to the data acquisition period of the actual data; generating a plurality of future prediction models that predict fluctuations in the actual data after the data acquisition period based on each of the plurality of extension data and the actual data; selecting a future prediction model with relatively high prediction accuracy from the plurality of future prediction models; and predicting the actual data after the data acquisition period using the extension data used to generate the future prediction model with relatively high prediction accuracy.

Images