JP2022091019A - Information processing apparatus, information processing method, and program - Google Patents

Abstract

To provide a mechanism to obtain appropriate inference results according to characteristics of explanatory variables.SOLUTION: An information processing system is configured: to accept the setting of a group column among training data composed of a plurality of columns; to generate a trained model for each value of the accepted group column; and to perform inference processing on new data by using the trained model corresponding to the value of the group column in the new data to be inferred among the generated trained models.SELECTED DRAWING: Figure 1

Description

The present invention relates to an information processing apparatus, an information processing method, and a program.

In recent years, a mechanism using a trained model generated by using deep learning has been utilized in various situations. For example, it is used for detecting objects from images, classifying objects, and is also used in the fields of speech recognition and language processing. Further, as disclosed in Patent Document 1, a method using a trained model is also used in a situation where a forecast value is calculated from various explanatory variables such as past demand results and weather, such as demand forecast. ing.

Japanese Unexamined Patent Publication No. 2020-115276

There are numerical variables and categorical variables as explanatory variables, but depending on the categorical variable, the influence of other explanatory variables on the prediction result may be completely different. For example, if the "product category" is "beverages", the "temperature" may affect the "number of products sold", but if the "product category" is "PC supplies", the "product" of the "temperature" It is considered that there is not much influence on the number of sales. In such a case, even if one model is trained with the entire training data, it is conceivable that appropriate modeling cannot be performed and the accuracy of prediction becomes low.

Thus, in order to obtain appropriate inference results, it may be necessary to train a plurality of different models depending on the value of the categorical variable.

Therefore, it is an object of the present invention to provide a mechanism for obtaining an appropriate inference result according to the characteristics of the explanatory variables.

The information processing system of the present invention is a receiving means that accepts the setting of a group column among learning data composed of a plurality of columns, and a generating means that generates a trained model for each value of the group column received by the receiving means. And, among the trained models generated by the generation means, the inference means for executing the inference processing for the new data by using the trained model corresponding to the value of the group column in the new data to be inferred is provided. It is characterized by that.

According to the present invention, it is possible to obtain an appropriate inference result according to the characteristics of the explanatory variables.

It is a figure which shows an example of the structure of an information processing system. It is a block diagram which shows an example of the hardware composition of an information processing apparatus 101. Flowchart showing processing during learning Flowchart showing processing at the time of inference Diagram for explaining the processing contents at the time of learning Diagram for explaining the processing contents at the time of inference The figure which shows the setting screen of various parameters

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a diagram showing an example of a system configuration according to an embodiment of the present invention.

The information processing system of the present invention is a system in which an information processing device 101 and an external device 102 are communicably connected via a network 110.

The information processing device 101 is a device operated by a user, and receives an execution instruction of various parameter settings, learning processing (processing of the flowchart of FIG. 3), and inference processing (processing of the flowchart of FIG. 4). Execute the process.

The external device 102 manages learning data and the like.

In this embodiment, the information processing device 101 will be described as executing the processing of the flowcharts of FIGS. 3 and 4, but the external device 102 may execute the processing. Further, the information processing apparatus 101 may perform the processing described as being performed by the external device 102, such as the management of learning data.

FIG. 2 is a block diagram showing an example of the hardware configuration of the information processing device 101 and the external device 102 of the present invention.

As shown in FIG. 2, the information processing device 101 and the external device 102 are CPU (Central Processing Unit) 201, ROM (Read Only Memory) 202, RAM (Random Access Memory) 203, and storage device 204 via the system bus 200. , Input controller 205, voice controller 206, video controller 207, memory controller 208, and communication I / F controller 209 are connected.

The CPU 201 comprehensively controls each device and controller connected to the system bus 200.

The ROM 202 or the external memory 213 is a control program executed by the CPU 201 such as a BIOS (Basic Input / Output System) or an OS (Operating System), a computer-readable program for realizing this information processing method, and various necessary programs. Holds data (including data table).

The RAM 203 functions as a main memory, a work area, and the like of the CPU 201. The CPU 201 realizes various operations by loading a program or the like necessary for executing the process from the ROM 202 or the external memory 213 into the RAM 203 and executing the loaded program.

The input controller 205 controls input from an input device such as a keyboard 210 or a pointing device such as a mouse (not shown). When the input device is a touch panel, the user can give various instructions by pressing (touching with a finger or the like) the icon, the cursor, or the button displayed on the touch panel.

Further, the touch panel may be a touch panel such as a multi-touch screen that can detect a position touched by a plurality of fingers.

The video controller 207 controls the display on an external output device such as the display 212. The display shall include the display of a notebook computer integrated with the main body. The external output device is not limited to the display, and may be, for example, a projector. Further, as for the device capable of accepting the above-mentioned touch operation, an input device is also provided.

The video controller 207 can control a video memory (VRAM) for display control, can use a part of the RAM 203 as a video memory area, or can separately provide a dedicated video memory. It is possible.

The memory controller 208 controls access to the external memory 213. External memory is connected to an external storage device (hard disk), flexible disk (FD), or PCMCIA card slot that stores boot programs, various applications, font data, user files, edit files, various data, etc. via an adapter. Compact flash (registered trademark) memory etc. can be used.

The communication I / F controller 209 connects to and communicates with an external device via a network, and executes communication control processing on the network. For example, communication using TCP / IP, telephone lines such as ISDN, and communication using 4G lines and 5G lines of mobile phones are possible.

The CPU 201 enables display on the display 212 by, for example, executing an outline font expansion (rasterization) process in the display information area in the RAM 203. Further, the CPU 201 enables a user instruction with a mouse cursor or the like (not shown) on the display 212.

Next, the process at the time of learning will be described with reference to the flowchart of FIG. 3 and the image diagram of FIG.

First, the whole picture of the processing at the time of learning will be described with reference to the image diagram of FIG.

The data shown in "all records" in FIG. 5 is the data that is the source of the learning data. As shown in the list of "all records", it is assumed that the objective variable is calculated using the data of the explanatory variables 1 to N and the data indicating the group (group column) as the explanatory variables. That is, the learning process is executed by associating the value of the "objective variable" with the "data indicating the explanatory variables 1 to N and the data indicating the group" as correct answer data.

Here, it is assumed that the group column is data (category variable) selected from predetermined unique values such as "A" or "B". On the other hand, the explanatory variables 1 to N may be data (values) (numerical variables) without any restrictions or categorical variables.

In such data, the trained model A learned by dividing the data into the data in which the value of the group column is "A" and the data in which the value in the group column is "A", and the group. Each trained model B trained using the data whose column value is "B" is generated.

Further, when the group is divided, the trained model for each unique value is not generated for the unique value in which the number of records is less than the preset minimum number of records. Specifically, when the minimum number of records is 10, the number of records whose group column value is "A" is 20, the number of records of "B" is 15, the number of records of "C" is 8, and "D". When the number of records is 5, trained models A and B are generated for the unique values "A" and "B" having the number of records satisfying 10.

Next, the process at the time of learning will be described using the flowchart of FIG.

In step S301, the input of the minimum number of records to be learned is accepted. A trained model is generated for each unique value among the values in the group column, for which the unique value has the number of records satisfying this minimum number of records to be learned.

In step S302, the setting of the name of the column to be the group column is accepted.

In step S303, the learning data is read. Specifically, it is the data shown by "all records" in FIG. 5, and is composed of an explanatory variable, a group column, and an objective variable.

In step S304, a list of unique values that can be set in the group column is acquired. For example, in the example of FIG. 5, "A" and "B" are unique values.

In step S305, the first unique value in the unique value list acquired in step S304 is acquired.

In step S306, among the training data acquired in step S303, the value of the group column is the unique value of the processing target (for the first processing, the first unique value of the list acquired in step S305. The second and subsequent times are The record group which is the unique value acquired in step S311) is acquired.

In step S307, it is determined whether the number of records in the record group acquired in step S306 satisfies the minimum number of records accepted in step S301.

When the minimum number of records is satisfied (step S307: YES), the process proceeds to step S308.

If not satisfied (step S307: NO), the process proceeds to step S310.

In step S308, the model is trained using the record group acquired in step S306 as training data, and a trained model is generated.

In step S309, the trained model generated in step S308 is saved as a trained model corresponding to the unique value to be processed.

In step S310, it is determined whether or not the processes of steps S306 to S309 have been executed for all the unique values included in the list acquired in step S304.

When the process is executed for all the unique values, the process proceeds to step S311.

If a unique value for which processing has not been executed remains, the processing proceeds to step S312.

In step S311, a trained model is generated using all the records of the training data acquired in step S303. Then, the generated trained model is saved in the same manner as the trained model for each other unique value.

In this embodiment, it has been described that the process of generating a trained model using all the records is executed in any case, but it is determined whether there is a unique value that does not satisfy the minimum number of records, and it is determined that the unique value exists. You may execute the process of generating the trained model using all the records only when it is done.

In step S312, the next unique value in the list acquired in step S304 is acquired. Then, the processes after step S306 are executed for the unique value.

By the above processing, a trained model for each value of the group column can be generated.

Next, the processing at the time of inference will be described with reference to the flowchart of FIG. 4 and the image diagram of FIG.

First, the whole picture of the processing at the time of inference will be described with reference to the image diagram of FIG.

The data shown by the new records 1 and 2 in FIG. 6 are the data to be inferred. A new record is data consisting of explanatory variables and group columns. The inference process is a process of calculating the objective variable by putting this data into the trained model.

For example, for the new record 1, since the group column is "B", the inference process is performed by the trained model B generated by learning using the record group whose group column is "B". The value "14" obtained by the inference process is the calculated (predicted) objective variable.

In this way, by performing inference using an appropriate trained model (learned using the record group of the value) according to the value of the group column, an inference result that appropriately considers the value of the group column is obtained. It becomes possible.

Next, the content of the inference process will be described with reference to the flowchart of FIG.

In step S401, a record to be predicted (a new record in FIG. 6) is acquired.

In step S402, the value of the group column of the prediction target record acquired in step S401 is acquired. ("B" in the new record 1 in FIG. 6).

In step S403, it is determined whether there is a trained model corresponding to the value of the group column acquired in step S402 (whether it exists in the trained model generated in step S308 and stored in S309).

If present (step S403: YES), the process proceeds to step S404.

If it does not exist (step S403: NO), the process proceeds to step S406.

In step S404, the trained model corresponding to the value of the group column acquired in step S402 is read.

In step S405, the inference process for the record acquired in step S401 is executed using the trained model read in step S404.

In step S406, it is notified that the trained model corresponding to the value of the group column acquired in step S402 does not exist.

In step S407, the trained model trained using all the records generated in step S311 is read.

In step S408, the inference process for the record acquired in step S401 is executed using the trained model read in step S407.

By the above processing, it becomes possible to perform inference using an appropriate trained model (learned using the record group of the value) according to the value of the group column, and as a result, the value of the group column is appropriately set. It is possible to obtain the inference result in consideration.

In addition, for the value of the group column whose number of records does not satisfy the minimum number of records, even if a trained model is generated only with the record of the value, a model that can be appropriately inferred is generated because the number of training data is small. Can not. Therefore, it is said that the inference process cannot be executed by generating a trained model using all the records and inferring the new data of the value of the group column whose number of records does not satisfy the minimum number of records by the trained model. It is possible to avoid the situation.

FIG. 7 is an example of a setting screen that accepts inputs such as the minimum number of records to be learned and the name of the group column. The input field shown in 701 is an input field that accepts the input of the minimum number of records to be learned. The input field of 702 is an input field that accepts the setting of the name of the column to be the group column.

Although the embodiment in which the processing is executed in the information processing apparatus 101 has been described above, the present invention can take an embodiment as, for example, a system, an apparatus, a method, a program, a recording medium, or the like. Specifically, it may be applied to a system composed of a plurality of devices, or may be applied to a device composed of one device.

Further, the program in the present invention is a program in which a computer can execute the processing method of the flowchart shown in FIGS. 3 and 4, and the storage medium of the present invention stores a program in which the computer can execute the processing method in FIGS. 3 and 4. Has been done.

As described above, a recording medium recording a program that realizes the functions of the above-described embodiment is supplied to the system or device, and the computer (or CPU or MPU) of the system or device stores the program in the recording medium. Needless to say, the object of the present invention can be achieved by reading and executing.

In this case, the program itself read from the recording medium realizes the novel function of the present invention, and the recording medium on which the program is recorded constitutes the present invention.

Recording media for supplying programs include, for example, flexible disks, hard disks, optical disks, optical magnetic disks, CD-ROMs, CD-Rs, DVD-ROMs, magnetic tapes, non-volatile memory cards, ROMs, EEPROMs, and silicon. A disc or the like can be used.

Further, by executing the program read by the computer, not only the function of the above-described embodiment is realized, but also the OS (operating system) or the like running on the computer is actually realized based on the instruction of the program. Needless to say, there are cases where a part or all of the processing is performed and the processing realizes the functions of the above-described embodiment.

Further, after the program read from the recording medium is written in the memory provided in the function expansion board inserted in the computer or the function expansion unit connected to the computer, the function expansion board is based on the instruction of the program code. It goes without saying that there are cases where the CPU or the like provided in the function expansion unit performs a part or all of the actual processing, and the processing realizes the functions of the above-described embodiment.

Further, the present invention may be applied to a system composed of a plurality of devices or a device composed of one device. It goes without saying that the present invention can also be applied when it is achieved by supplying a program to a system or an apparatus. In this case, by reading the recording medium containing the program for achieving the present invention into the system or the device, the system or the device can enjoy the effect of the present invention.

Further, by downloading and reading a program for achieving the present invention from a server, database, or the like on a network by a communication program, the system or device can enjoy the effect of the present invention. It should be noted that the present invention also includes all the configurations in which each of the above-described embodiments and modifications thereof are combined.

101 Information processing device 102 External device 110 Network

Claims (6)

A reception means that accepts group column settings among learning data composed of multiple columns,
A generation means for generating a trained model for each value of the group column received by the reception means, and a generation means.
Among the trained models generated by the generation means, the inference means for executing the inference processing for the new data using the trained model corresponding to the value of the group column in the new data to be inferred.
An information processing system characterized by being equipped with. The receiving means further accepts a minimum number of records for determining whether to generate a trained model for each value of the group column.
The generation means is characterized in that, among the values of the group column in the training data, the trained model for each value of the group column is generated for the value in which the number of records satisfies the minimum number of records accepted by the reception means. The information processing system according to claim 1. The generation means further comprises generating a trained model using all the records of the training data.
When the trained model corresponding to the value of the group column in the new data to be inferred is not generated, the inference means uses the trained model generated by using all the records of the training data to generate the new data. The information processing system according to claim, which comprises performing inference processing on the data. The information processing system according to claim 2 or 3, further comprising a notification means for notifying that a trained model corresponding to a value of a group column in the new data to be inferred has not been generated. The reception process in which the reception means of the information processing system accepts the setting of the group column among the learning data composed of a plurality of columns,
A generation step in which the generation means of the information processing system generates a trained model for each value of the group column received by the reception step.
The inference means of the information processing system executes inference processing on the new data using the trained model corresponding to the value of the group column in the new data to be inferred among the trained models generated by the generation step. Inference process and
An information processing method characterized by being provided with. Computer,
A reception means that accepts group column settings among learning data composed of multiple columns,
A generation means for generating a trained model for each value of the group column received by the reception means, and a generation means.
Of the trained models generated by the generation means, a program for functioning as an inference means for executing inference processing on the new data using the trained model corresponding to the value of the group column in the new data to be inferred. ..

Images