JP2022029347A - Feature selection program, device, and method - Google Patents

Abstract

To provide a feature amount selection device capable of selecting feature quantity for effective machine learning from items included in a data set without requiring correct data.SOLUTION: The feature amount selection device acquires a shaped data set and a formatting processing history DB21 storing formatting processing history data. Using a formatting function stored as formatting processing history data as a key, a "formatting function" in a priority item DB22, in which priority items that should be selected is given to the feature quantity to be included in the training data, is searched, and corresponding to "priority item target argument" is acquired from the formatting processing history data. An item corresponding to the acquired priority item target argument, which is used as an argument when executing the formatting function, is extracted as the priority item, and the priority item is selected as the feature quantity.SELECTED DRAWING: Figure 11

Description

The disclosed technique relates to a feature amount selection program, a feature amount selection device, and a feature amount selection method.

Using a model generated by collecting multiple types of sensor data indicating the operating status of production equipment in a factory and machine learning the collected sensor data data set, abnormalities in production equipment, etc. without human intervention. Alternatively, the introduction of a system for detecting normality is progressing. The collected data set often contains a large number of items, and if the data set is used as it is for machine learning, there is a problem that machine learning takes time. In addition, the model becomes complicated, and there arises a problem that the interpretability at the time of inference using the model is lowered.

Therefore, a technique has been proposed in which a predetermined number of items are selected as feature quantities to be included in the learning data used for machine learning from a large number of items included in the data set. For example, an input variable selection support device has been proposed that supports input variable selection by presenting an index for narrowing down input variables from input candidate variables by a simple method. This device uses model information data standardized so that the mean value is 0 and the standard deviation is 1, and multiple input / output models that represent the relationship between all input candidate variables and output variables can be created using different model principles. Generate over multiple formats. Then, this device performs sensitivity generation for each input / output model to generate the sensitivity which is the absolute value of the output variable for each input candidate variable of each input / output model, and inputs the input variable based on the sensitivity of each input / output model. Select and present.

In addition, a selection device that selects a variable that is in the range of values that the data group from the observation target can take has been proposed. This device clusters variables based on similarity and selects a specific representative variable from the variables belonging to each cluster. Then, based on the knowledge of a skilled manager or technician, this device sets a label that specifies a variable to be monitored from the input device and stores it in the label specification table.

There is also a method of selecting features based on the nature of the data in the dataset without using a machine-learned model. Specifically, there is a method of selecting a feature amount based on the relationship between the objective variable and the explanatory variable. Alternatively, in order to avoid a decrease in model accuracy due to multicollinearity, there is a method of deleting items having a high correlation between explanatory variables.

Japanese Unexamined Patent Publication No. 2010-282547 International Publication No. 2018/092317

"Summary of feature selection", [online], July 8, 2020, [Search on July 10, 2020], Internet <URL: https://qiita.com/shimopino/items/5fee7504c7acf044a521>

In the prior art, an objective variable (output variable), that is, correct answer data is required to select a feature amount. However, for example, in a system for detecting an abnormality in a machine as described above, there may be no information indicating an abnormality that is correct data in machine learning. Therefore, there is a problem that the conventional technique that requires correct answer data cannot be applied. In addition, it is desirable to select important business items as features without deleting them, but when deleting items with high correlation between explanatory variables, there is a possibility that the important explanatory variables will be deleted. be.

As one aspect, the disclosed technique aims to select features for effective machine learning from the items contained in the dataset without the need for correct data.

As one embodiment, the disclosed technique defines a data set containing a plurality of feature quantities and for which shaping processing has been executed, a history of the shaping processing, and a specific feature quantity as an argument in the specific shaping process. Get specific information. Then, the disclosed technique selects a feature amount corresponding to the specific feature amount defined in the specific information from the plurality of feature amounts included in the data set based on the acquired history of the shaping process. do.

One aspect is that it is possible to select a feature amount for effective machine learning from the items included in the data set without requiring correct answer data.

It is a functional block diagram of a feature quantity selection device. It is a figure which shows an example of the data set A before the shaping process. It is a figure which shows an example of the data set B before the shaping process. It is a figure which shows an example of a shaping process screen. It is a figure which shows an example of the data set A which has been shaped. It is a figure which shows another example of a shaping process screen. It is a figure which shows an example of the data set B which has been shaped. It is a figure which shows an example of the shaping process history DB about the data set A. It is a figure which shows an example of the shaping process history DB about the data set B. It is a figure which shows an example of a priority item DB. It is a figure for demonstrating the extraction of a priority item. It is a figure which shows an example of the correlation table of a priority item and another item. It is a figure which shows an example of a reception screen. It is a figure for demonstrating the update of a priority item DB. It is a block diagram which shows the schematic structure of the computer which functions as a feature quantity selection device. It is a flowchart which shows an example of a feature amount selection process.

Hereinafter, an example of the embodiment according to the disclosed technology will be described with reference to the drawings.

As shown in FIG. 1, the feature amount selection device 10 functionally includes a shaping unit 11, an acquisition unit 12, an extraction unit 13, a selection unit 14, a reception unit 15, and an update unit 16. .. The extraction unit 13 and the selection unit 14 are examples of the “selection unit” of the disclosed technology.

The shaping unit 11 acquires the data set before the shaping process input to the feature amount selection device 10. 2 and 3 show an example of the data set before the shaping process. Hereinafter, the data set shown in FIG. 2 will be referred to as “data set A”, and the data set shown in FIG. 3 will be referred to as “data set B”. The data set contains a plurality of items, and each data has a value for each item. As shown in FIG. 2, the data set A includes items "X1", "X2", "X3", "X4", and "X5". Further, as shown in FIG. 3, the data set B includes the items "operation start date", "failure occurrence date", "X1", "X2", and "X3".

The shaping unit 11 executes the shaping process on the acquired data set before the shaping process. Specifically, the shaping unit 11 is a function executed as a shaping process for the data set (hereinafter referred to as "shaping process function"), and a shaping processing function among a plurality of items included in the data set before the shaping process. Accepts the specification of the item that becomes the argument of. Then, the shaping unit 11 executes the designated shaping processing function with the designated item as an argument.

As the shaping processing function, a plurality of predetermined types of shaping processing functions can be prepared. The formatting function is, for example, a numerical type check that checks whether the value of the specified item is a numerical value, or a numerical type value check that checks whether the value of the specified item is a value in the specified range. And so on. Further, the formatting function is, for example, a date / time type conversion that converts a value of a date / time type item into a date / time type value of another format, and a difference between the values of two specified date / time type items of a predetermined date / time type. Date and time type difference conversion to convert to a value. Further, the shaping processing function is, for example, a category type conversion that converts the value of the specified item into a value of a different category, a defect that complements the missing value among the values of the specified item by prediction processing, or the like. Value complementation, etc.

For example, the shaping unit 11 displays the shaping processing screen 30 as shown in FIG. 4 on the display device. In the example of FIG. 4, the shaping process screen 30 has a file name area 31 in which the file name of the data set to be shaped is displayed, an execution area 32 for designating the content and execution of the shaping process, and the shaping process. The result area 33 in which the result of is displayed is included. In the execution area 32, a plurality of predetermined types of shaping processing functions are displayed in a list in a selectable state. Further, in the execution area 32, when a shaping process function is selected from the list, a text box for specifying an item to be an argument of the shaping process function and an execution button for instructing the execution of the shaping process are displayed. Will be done.

For example, the person in charge of the shaping process reads the file of the target data set, selects the shaping processing function to be executed from the list in the execution area 32 of the shaping process screen 30, and inputs the item to be an argument in the text box. And press the execute button. The shaping unit 11 acquires information on the shaping processing function specified on the shaping processing screen 30 and items to be arguments, and executes the shaping process based on the acquired information on the data set. Then, the shaping unit 11 displays the execution result of the shaping process in the result area 33 of the shaping process screen 30.

FIG. 4 shows an example of executing the formatting process of the numerical type check for the data set A. Note that FIG. 4 shows the shaping processing function selected by the check mark. The same applies to FIG. 6 described later. Further, as an argument, "X1" is specified as an item for checking the numerical type. In this example, the value in the third row of the item "X1" is "0.1", and the result indicating that it is not a numerical value is displayed in the result area 33. In this example, the shaping unit 11 may automatically correct the value based on the check result, or may accept the correction of the corresponding value by the person in charge. FIG. 5 shows the data set A after the shaping process including the correction of the value.

FIG. 6 shows another example of the shaping processing screen 30. FIG. 6 shows an example of executing the shaping process of the datetime type difference conversion for the data set B. As arguments, "operation start date" is specified for item name 1, "failure occurrence date" is specified for item name 2, "operating period" is specified for the difference conversion result item name, and "day" is specified for the difference result date and time type. There is. This indicates that the shaping process of adding the difference number of days between the operation start date and the failure occurrence date as an item called the operation period is executed. Therefore, the shaping unit 11 adds the result of calculating the period from the operation start date to the failure occurrence date on a daily basis, for example, as the item name “operating period” in the last column of the data set. FIG. 7 shows the data set B after the shaping process.

The shaping unit 11 stores the history of the shaping process executed on the data set in the shaping process history DB 21. In the shaping process history DB 21, the executed shaping processing function and the items as its arguments are stored as shaping processing history data. When a plurality of shaping processing functions are executed for one data set, the shaping unit 11 stores a plurality of shaping processing history data in the shaping processing history DB 21.

FIG. 8 shows an example of the shaping process history DB 21 that stores the history of shaping processing for the data set A. As shown in FIG. 8, in the shaping process history DB 21, the argument name of the shaping processing function and the information of the item that became the argument correspond to the "shaping processing function" executed for the data set as "arguments". Attached and memorized. FIG. 8 shows an example in which the shaping process of the numerical type value check for the item “X3” is executed in addition to the numerical type check described in the example of the shaping processing screen 30 of FIG. Further, FIG. 9 shows an example of the shaping process history DB 21 that stores the history of the shaping process for the data set B. FIG. 9 shows an example in which the shaping process of the datetime type difference conversion described in the example of the shaping process screen 30 of FIG. 6 is executed.

The acquisition unit 12 acquires a data set that has been shaped from the shaping unit 11. In addition, the acquisition unit 12 acquires the shaping process history DB 21 for the acquired data set that has been shaped, from the shaping process history DB 21. Further, the acquisition unit 12 acquires the priority item DB 22 stored in the external storage device. Among the plurality of types of shaping processing functions, the priority item DB 22 defines a specific item as an argument in the specific shaping processing function. The specific item that becomes an argument in the specific shaping processing function is an item that should be preferentially selected as a feature amount to be included in the training data (hereinafter referred to as "priority item"). Specifically, an item that has undergone shaping processing on the assumption that the value of that item will be used after the shaping process is defined as a priority item. The information stored in the priority item DB 22 is an example of "specific information" of the disclosed technology. FIG. 10 shows an example of the priority item DB 22. In the example of FIG. 10, among the items that are arguments in the shaping process function, the argument that is the priority item is stored in association with the "shaping process function" that indicates the specific shaping process as the "priority item target argument". There is. The acquisition unit 12 passes the acquired information to the extraction unit 13.

Based on the shaping process history DB 21 passed from the acquisition unit 12, the extraction unit 13 is an item corresponding to the priority item target argument defined in the priority item DB 22 from a plurality of items included in the formatted data set. Is extracted as a priority item in the data set.

Specifically, as shown in FIG. 11, the extraction unit 13 searches for the "shaping function" of the priority item DB 22 using the shaping processing function stored in the shaping processing history DB 21 as the shaping processing history data as a key. .. The extraction unit 13 acquires the "priority item target argument" associated with the "shaping process function" that matches the shaping process function used as the key in the priority item DB 22. Then, the extraction unit 13 extracts the item that became an argument at the time of execution of the shaping processing function corresponding to the acquired priority item target argument from the shaping processing history data as a priority item. In the example of the data set A, as shown in FIG. 11, the “check item name” is acquired as the priority item target argument, and the item “X1” corresponding to the check item name is extracted as the priority item.

The extraction unit 13 performs the above processing on all of the shaping processing history data stored in the shaping processing history DB 21. Therefore, if each of the plurality of shaping processing functions included in the shaping processing history data matches any of the "shaping processing functions" of the priority item DB 22, the plurality of priority items will be extracted. In the example of FIG. 11, since the shaping processing function "numerical type value check" included in the shaping processing history data is not defined in the priority item DB 22, the corresponding item "X3" is not extracted as a priority item. The extraction unit 13 passes the extracted priority item information and the formatted data set to the selection unit 14.

The selection unit 14 calculates an index showing the correlation between the priority item passed from the extraction unit 13 and each of the other items other than the priority item among the plurality of items included in the formatted data set. .. The selection unit 14 creates a correlation table showing the correlation between the priority item and other items, as shown in FIG. 12, for example. In the example of FIG. 12, the correlation table stores a correlation coefficient calculated as an index showing the correlation between the priority item and other items.

The selection unit 14 first selects a priority item as a feature amount to be included in the learning data. Then, the selection unit 14 selects another item whose index indicating the correlation with the priority item is lower than the predetermined threshold value as the feature amount to be included in the learning data. The reason for selecting other items that have a low correlation with the priority item is to avoid the occurrence of multicollinearity with the priority item during machine learning. The threshold value may be a predetermined value, or the Nth smallest index in the correlation table may be used as the threshold value. In the latter case, N other items will be selected as features. In the example of FIG. 12, when the threshold value is 0.7, the selection unit 14 selects the items “X1”, “X2”, and “X5” as feature quantities. The selection unit 14 passes information on a plurality of items included in the data set and items selected as feature quantities to the reception unit 15.

The reception unit 15 presents each of the plurality of items included in the data set to, for example, a person in charge of model design, together with information indicating whether or not the items are selected as the feature amount. Then, the reception unit 15 receives from the person in charge the addition or cancellation of the selection as the feature amount for any of the items.

For example, the reception unit 15 displays the reception screen 35 as shown in FIG. 13 on the display device. In the example of FIG. 13, the reception screen 35 has a file name area 36 in which the file name of the data set after the shaping process is displayed, and a reception area for instructing addition or cancellation of feature quantity selection and confirmation. 37 and are included. In the reception area 37, whether or not each of the plurality of items is selected as a feature amount is displayed in a state in which it can be corrected. Further, in the reception area 37, a confirmation button pressed when confirming the selection of the feature amount is displayed. In the example of FIG. 13, a check mark is displayed in association with the item selected as the feature amount.

When adding the selection of the feature amount, the person in charge adds a check mark to the corresponding item as shown by the broken line portion in the lower part of FIG. In addition, the person in charge cancels the check mark of the corresponding item when deselecting the feature amount. When the person in charge finishes modifying the selection of the feature amount, he / she presses the confirm button. As a result, the reception unit 15 receives and outputs the confirmation information of the feature amount selection. Further, the reception unit 15 updates the information of the items selected as the feature amount other than the items selected by the selection unit 14, that is, the items to which the person in charge has added the selection as the feature amount. Hand over to. In the example of FIG. 13, the items "X1", "X2", "X3", and "X5" are output as the confirmation information of the feature amount selection, and the information of the item "X3" is updated as the item to which the selection is added. It is handed over to department 16.

The update unit 16 adds a shaping processing function having an item whose selection has been added by the person in charge as an argument to the priority item DB 22. Specifically, as shown in FIG. 14, the update unit 16 searches for the "argument" of the shaping processing history data for the corresponding data set using the added item as a key, and the added item becomes an argument. Extract the included formatting functions. The update unit 16 searches for the "formatting function" of the priority item DB 22 using the extracted shaping processing function as a key. Then, when the shaping processing function used as a key is not registered in the priority item DB 22, the update unit 16 adds the shaping processing function and the priority item target argument to the priority item DB 22. The priority item target argument is an item for which a selection has been added by the person in charge. In the example of FIG. 14, the shaping processing function “numeric type value check” in which the added item “X3” is used as an argument “check item name” is extracted. Since the "numeric type value check" is not registered in the priority item DB 22 before the update, the "numeric type value check" and the priority item target argument "check item name" are added to the priority item DB 22.

The feature amount selection device 10 can be realized by, for example, the computer 40 shown in FIG. The computer 40 includes a CPU (Central Processing Unit) 41, a memory 42 as a temporary storage area, and a non-volatile storage unit 43. Further, the computer 40 includes an input / output device 44 such as an input unit and a display unit, and an R / W (Read / Write) unit 45 that controls reading and writing of data to the storage medium 49. Further, the computer 40 includes a communication I / F (Interface) 46 connected to a network such as the Internet. The CPU 41, the memory 42, the storage unit 43, the input / output device 44, the R / W unit 45, and the communication I / F 46 are connected to each other via the bus 47.

The storage unit 43 can be realized by an HDD (Hard Disk Drive), an SSD (Solid State Drive), a flash memory, or the like. In the storage unit 43 as a storage medium, a feature amount selection program 50 for making the computer 40 function as the feature amount selection device 10 is stored. The feature amount selection program 50 includes a shaping process 51, an acquisition process 52, an extraction process 53, a selection process 54, a reception process 55, and an update process 56.

The CPU 41 reads the feature amount selection program 50 from the storage unit 43, expands the feature amount selection program 50 into the memory 42, and sequentially executes the processes included in the feature amount selection program 50. The CPU 41 operates as the shaping unit 11 shown in FIG. 1 by executing the shaping process 51. Further, the CPU 41 operates as the acquisition unit 12 shown in FIG. 1 by executing the acquisition process 52. Further, the CPU 41 operates as the extraction unit 13 shown in FIG. 1 by executing the extraction process 53. Further, the CPU 41 operates as the selection unit 14 shown in FIG. 1 by executing the selection process 54. Further, the CPU 41 operates as the reception unit 15 shown in FIG. 1 by executing the reception process 55. Further, the CPU 41 operates as the update unit 16 shown in FIG. 1 by executing the update process 56. Further, the CPU 41 expands each of the shaping process history DB 21 and the priority item DB 22 into the memory 42 when each process is executed. As a result, the computer 40 that has executed the feature amount selection program 50 functions as the feature amount selection device 10. The CPU 41 that executes the program is hardware.

The function realized by the feature amount selection program 50 can also be realized by, for example, a semiconductor integrated circuit, more specifically, an ASIC (Application Specific Integrated Circuit) or the like.

Next, the operation of the feature amount selection device 10 according to the present embodiment will be described. When the feature amount selection process to be included in the learning data is instructed, the feature amount selection process shown in FIG. 16 is executed in the feature amount selection device 10. The feature amount selection process is an example of the feature amount selection method of the disclosed technology.

In step S11, the shaping unit 11 reads the data set before the shaping process.

Next, in step S12, the shaping unit 11 accepts the specification of the shaping processing function to be executed for the data set and the item to be the argument of the shaping processing function, and specifies the read data set before the shaping process. Executes the specified formatting function with the specified item as an argument. Then, the shaping process history data in which the shaping process function executed for the data set and the item as an argument thereof are associated with each other is stored in the shaping process history DB 21 by the shaping unit 11.

Next, in step S13, the acquisition unit 12 acquires the shaping processing data set from the shaping unit 11, and acquires the shaping processing history data for the acquired shaping processing data set from the shaping processing history DB 21. The priority item DB 22 stored in the external storage device is acquired. Then, the acquisition unit 12 passes the acquired information to the extraction unit 13. The extraction unit 13 is an item corresponding to the priority item target argument defined in the priority item DB 22 from a plurality of items included in the formatted data set based on the shaping processing history data passed from the acquisition unit 12. Is extracted as a priority item in the data set. The extraction unit 13 passes the extracted priority item information and the formatted data set to the selection unit 14.

Next, in step S14, the selection unit 14 sets the priority items passed from the extraction unit 13 and each of the plurality of items included in the formatted data set other than the priority items. An index showing the correlation (for example, a correlation coefficient) is calculated.

Next, in step S15, the selection unit 14 first selects a priority item as a feature amount to be included in the learning data. Further, the selection unit 14 selects another item whose index indicating the calculated correlation is lower than a predetermined threshold value as the feature amount to be included in the learning data. The selection unit 14 passes information on a plurality of items included in the data set and items selected as feature quantities to the reception unit 15.

Next, in step S16, the reception unit 15 displays on the display device a reception screen 35 in which it is possible to modify whether or not each of the plurality of items included in the data set is selected as a feature amount. .. Then, the reception unit 15 receives from the person in charge the addition or cancellation of the selection as the feature amount for any item, that is, the correction of the selection of the feature amount, via the reception screen 35.

Next, in step S17, the reception unit 15 receives the information of the item selected as the feature amount when the confirmation button is pressed on the reception screen 35, and outputs the information as the confirmation information of the feature amount selection. Further, the reception unit 15 passes the information of the item to which the person in charge has added the selection as the feature amount to the update unit 16.

Next, in step S18, if the shaping processing function having the item to which the selection has been added by the person in charge as an argument is not registered in the priority item DB 22, the updating unit 16 has the shaping processing function and the priority item target argument. Is added to the priority item DB22. As a result, the priority item DB 22 is updated, and the feature amount selection process ends.

As described above, the feature amount selection device according to the present embodiment executes the shaping process on the data set and stores the history of the shaping process. Then, the feature quantity selection device includes a plurality of shaped data sets included in the shaped data set based on the history of the shaping process and the priority item DB in which the specific item as an argument of the specific shaping process function is defined as the priority item. Priority items are extracted from the items of and selected as features. As a result, it is possible to select a feature amount for effective machine learning from the items included in the data set without requiring correct answer data.

Further, the feature amount selection device according to the present embodiment also selects other items having a low correlation with the priority item as the feature amount. This makes it possible to further select the features to be included in the training data while avoiding the occurrence of multicollinearity. When the number of feature quantities to be included in the learning data is sufficient only for the priority items, it is not always necessary to perform the process of selecting other items having a low correlation with the priority items as the feature quantities.

Further, the feature amount selection device according to the present embodiment accepts modifications to the selected feature amount and confirms the selection of the feature amount. Thereby, for example, the feature amount can be selected based on the judgment of the person in charge of model design or the like. It is not essential to accept the modification of the feature amount selection, and the feature amount selected by the selection unit may be output as the confirmation information of the feature amount selection.

Further, the feature amount selection device according to the present embodiment adds an item to which the person in charge has added selection as a feature amount and a shaping processing function having the item as an argument to the priority item DB. As a result, the priority item DB can be updated according to the actual situation of business and the like. It is not essential to update the priority item DB. Also, instead of updating the priority item DB each time the person in charge adds a selection as a feature amount, the priority item DB is updated when the selection is added a predetermined number of times for the same item. You may do so.

Further, in the above embodiment, the case where the feature amount selection device executes the shaping process for the data set has been described, but the present invention is not limited to this. For example, the shaping process may be executed by an external device and the shaping process history data may be stored. In this case, the configuration of the shaping unit may be omitted from the feature amount selection device, and the acquisition unit may read the data set having been shaped and the shaping processing history DB from the external device.

Further, the shaping processing function exemplified in the above embodiment is an example.

Further, in the above embodiment, the embodiment in which the feature amount selection program is stored (installed) in the storage unit in advance has been described, but the present invention is not limited to this. The program according to the disclosed technology can also be provided in a form stored in a storage medium such as a CD-ROM, a DVD-ROM, or a USB memory.

The following additional notes are further disclosed with respect to the above embodiments.

(Appendix 1)
A data set containing a plurality of feature quantities and for which shaping processing has been executed, a history of the shaping processing, and specific information in which a specific feature quantity as an argument in a specific shaping process is defined are acquired.
Based on the acquired history of the shaping process, a computer performs a process including selecting a feature amount corresponding to the specific feature amount defined in the specific information from the plurality of feature amounts included in the data set. Feature selection program to be executed by.

(Appendix 2)
Each of the first feature amount selected as the feature amount corresponding to the specific feature amount and the second feature amount other than the first feature amount among the plurality of feature amounts included in the data set. The feature amount selection program according to Appendix 1, which calculates an index showing a correlation with and selects the second feature amount whose index is lower than a predetermined threshold value together with the first feature amount.

(Appendix 3)
An appendix for presenting each of the plurality of feature quantities to the user together with information indicating whether or not the feature quantity has been selected, and causing the computer to perform a process further including accepting addition or cancellation of selection for any of the feature quantities. The feature amount selection program described in 1 or Appendix 2.

(Appendix 4)
The feature amount selection program according to Appendix 3 for causing the computer to perform a process including adding to the specific information a shaping process using a feature amount to which a selection has been added by a user as an argument.

(Appendix 5)
Any one of Supplementary note 1 to Supplementary note 4 for executing a shaping process on the data set before the shaping process and further causing the computer to perform a process including storing the history of the shaping process in a predetermined storage unit. Feature selection program described in section.

(Appendix 6)
A data set that includes a plurality of feature quantities and has been subjected to shaping processing, a history of the shaping processing, and an acquisition unit that acquires specific information in which a specific feature quantity that is an argument in a specific shaping process is defined. ,
A selection unit that selects a feature amount corresponding to the specific feature amount defined in the specific information from the plurality of feature amounts included in the data set based on the acquired history of the shaping process.
Feature quantity selection device including.

(Appendix 7)
The selection unit includes a first feature amount selected as a feature amount corresponding to the specific feature amount, and a second feature amount other than the first feature amount among the plurality of feature amounts included in the data set. The feature amount selection according to Appendix 6, which calculates an index showing the correlation with each of the feature amounts of, and selects the second feature amount whose index is lower than a predetermined threshold value together with the first feature amount. Device.

(Appendix 8)
The feature according to Appendix 6 or Appendix 7, further including a reception unit that presents each of the plurality of feature quantities to the user together with information indicating whether or not the feature quantity has been selected, and accepts addition or cancellation of selection for any of the feature quantities. Quantity selection device.

(Appendix 9)
The feature amount selection device according to Appendix 8, further including an update unit for adding a shaping process using a feature amount to which a selection has been made by a user as an argument to the specific information.

(Appendix 10)
The feature amount selection device according to any one of Supplementary note 6 to Supplementary note 9, which further includes a shaping unit that executes shaping processing on the data set before shaping processing and stores the history of the shaping process in a predetermined storage unit. ..

(Appendix 11)
A data set containing a plurality of feature quantities and for which shaping processing has been executed, a history of the shaping processing, and specific information in which a specific feature quantity as an argument in a specific shaping process is defined are acquired.
Based on the acquired history of the shaping process, a computer performs a process including selecting a feature amount corresponding to the specific feature amount defined in the specific information from the plurality of feature amounts included in the data set. Feature selection method to be executed by.

(Appendix 12)
Each of the first feature amount selected as the feature amount corresponding to the specific feature amount and the second feature amount other than the first feature amount among the plurality of feature amounts included in the data set. The feature amount selection method according to Appendix 11, wherein an index showing a correlation with the above is calculated, and the second feature amount whose index is lower than a predetermined threshold is selected together with the first feature amount.

(Appendix 13)
Appendix 11 or, the computer performs a process including presenting each of the plurality of feature quantities to the user together with information indicating whether or not the feature quantity has been selected, and accepting addition or cancellation of selection for any of the feature quantities. The feature amount selection method according to Appendix 12.

(Appendix 14)
The feature amount selection method according to Appendix 13, wherein the computer further executes a shaping process including adding a feature amount to which a selection has been added by a user as an argument to the specific information.

(Appendix 15)
Item 1. Described feature quantity selection method.

(Appendix 16)
A data set containing a plurality of feature quantities and for which shaping processing has been executed, a history of the shaping processing, and specific information in which a specific feature quantity as an argument in a specific shaping process is defined are acquired.
Based on the acquired history of the shaping process, a computer performs a process including selecting a feature amount corresponding to the specific feature amount defined in the specific information from the plurality of feature amounts included in the data set. A storage medium that stores a feature selection program to be executed by a computer.

10 Feature amount selection device 11 Shaped unit 12 Acquisition unit 13 Extraction unit 14 Selection unit 15 Reception unit 16 Update unit 21 Formatting processing history DB
22 Priority item DB
30 Formatting screen 35 Reception screen 40 Computer 41 CPU
42 Memory 43 Storage unit 49 Storage medium 50 Feature quantity selection program

Claims (7)

A data set containing a plurality of feature quantities and for which shaping processing has been executed, a history of the shaping processing, and specific information in which a specific feature quantity as an argument in a specific shaping process is defined are acquired.
Based on the acquired history of the shaping process, a computer performs a process including selecting a feature amount corresponding to the specific feature amount defined in the specific information from the plurality of feature amounts included in the data set. Feature selection program to be executed by. Each of the first feature amount selected as the feature amount corresponding to the specific feature amount and the second feature amount other than the first feature amount among the plurality of feature amounts included in the data set. The feature amount selection program according to claim 1, wherein an index showing a correlation with the above is calculated, and the second feature amount whose index is lower than a predetermined threshold value is selected together with the first feature amount. A claim for presenting each of the plurality of feature quantities to the user together with information indicating whether or not the feature quantity has been selected, and causing the computer to perform a process further including accepting addition or cancellation of selection for any of the feature quantities. The feature amount selection program according to claim 1 or claim 2. The feature amount selection program according to claim 3, wherein the computer is to perform a process including further adding to the specific information a shaping process having a feature amount to which a selection has been added by a user as an argument. Any of claims 1 to 4 for executing a shaping process on a data set before the shaping process and causing the computer to perform a process including further storing the history of the shaping process in a predetermined storage unit. The feature amount selection program described in item 1. A data set that includes a plurality of feature quantities and has been subjected to shaping processing, a history of the shaping processing, and an acquisition unit that acquires specific information in which a specific feature quantity that is an argument in a specific shaping process is defined. ,
A selection unit that selects a feature amount corresponding to the specific feature amount defined in the specific information from the plurality of feature amounts included in the data set based on the acquired history of the shaping process.
Feature quantity selection device including. A data set containing a plurality of feature quantities and for which shaping processing has been executed, a history of the shaping processing, and specific information in which a specific feature quantity as an argument in a specific shaping process is defined are acquired.
Based on the acquired history of the shaping process, a computer performs a process including selecting a feature amount corresponding to the specific feature amount defined in the specific information from the plurality of feature amounts included in the data set. Feature selection method to be executed by.

Images