Method of creation, calculation method, processing device, program, and storage medium

Abstract

To provide a creation method, a calculation method, a processing apparatus, a program, and a storage medium which can create a model which can calculate an estimated money amount in a more accurate manner.SOLUTION: A creation method according to an embodiment of the present invention has a step of converting requirements of an object of an estimation target into a first specification item and a second specification item. The creation method further has a step of, based on distribution of specification values of the first specification item and specification values of the second specification item in a plurality of instances relating to the object, extracting a part of the instances. The creation method yet further has a step of, with reference to related specification items having correlation with the object, creating an estimation model indicating a relation between money amounts in the part of the instances and the specification values o the related specification items.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Embodiments of the present invention relate to a creation method, a calculation method, a processing device, a program, and a storage medium.

Background Art

[0002] In some cases, a provider of services related to a property may be required to present an estimated amount to a customer before actually providing the service. There is a need for a technology that can calculate the estimated amount more accurately and more simply.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The problem to be solved by the present invention is to provide a creation method, a calculation method, a processing device, a program, and a storage medium that can create a model capable of calculating an estimated amount more accurately.

Means for Solving the Problems

[0005] The creation method according to the embodiment includes a step of converting requirements for a property to be estimated into a first specification item and a second specification item. The creation method further includes a step of extracting some of the cases based on the distributions of the specification values of the first specification item and the second specification item in a plurality of cases related to the property. The creation method further includes a step of referring to related specification items having a relevance to the property and creating an estimation model showing the relationship between the amount in some of the cases and the specification values of the related specification items.

Brief Description of the Drawings

[0006] [Figure 1]This is a flowchart showing a method for calculating an estimated amount, which includes a method for creating an embodiment of the present invention. [Figure 2] This is an example of quotation specification data. [Figure 3] This is an example of case data. [Figure 4] This is an example of a mapped distribution. [Figure 5] This is a schematic diagram illustrating an example of a method for extracting case studies. [Figure 6] This is a schematic diagram illustrating the estimation model that was created. [Figure 7] Here is another example of a mapped distribution. [Figure 8] This is a schematic diagram illustrating another example of a method for extracting case examples. [Figure 9] This is a schematic diagram showing a calculation system according to an embodiment. [Figure 10] This is a schematic diagram illustrating a user interface. [Figure 11] This is a schematic diagram illustrating a user interface. [Figure 12] This is a schematic diagram illustrating a method for creating an estimation model using confidence levels. [Figure 13] This is a schematic diagram illustrating how to create an estimation model using machine learning with confidence levels. [Figure 14] This is a schematic diagram showing the input data to the estimation model and the output data from the estimation model. [Figure 15] This flowchart shows a method for calculating an estimated price, which includes a method for creating a modified embodiment. [Figure 16] This is a schematic diagram showing the configuration of a calculation system according to a modified embodiment. [Figure 17] This is another example of quotation specification data. [Figure 18] This is an example of a classification master. [Figure 19] This is a schematic diagram representing the hardware configuration. [Modes for carrying out the invention]

[0007] The embodiments of the present invention will be described below with reference to the drawings. In this specification and in the drawings, elements similar to those already described are denoted by the same reference numerals, and detailed descriptions will be omitted as appropriate.

[0008] The invention according to this embodiment is used to create an estimation model that can more accurately estimate the amount of services related to an object. The object is movable property such as articles, or immovable property such as buildings such as power plants or office buildings. Services related to articles include the production and repair of articles. Services related to buildings include the construction and repair of buildings.

[0009] Regarding goods, if the goods provided are mass-produced, it is relatively easy to calculate the estimated cost of production or repair. On the other hand, for custom-made products, such as indented products, the specifications differ for each item, making it more difficult to calculate the estimated cost compared to mass-produced items. Indented products are large items such as power receiving and transforming equipment and power generation equipment. For custom-made products, since the specifications differ for each item, the estimated cost is calculated by considering the specific specifications each time production or repair is required. In order to calculate an accurate estimated cost, it is necessary to specifically identify the specifications, which is time-consuming. The invention according to this embodiment aims to calculate the estimated cost of custom-made products more accurately and easily. The following describes an example where the object is a goods.

[0010] (Calculation method) Figure 1 is a flowchart showing a method for calculating an estimated price using the creation method according to the embodiment. Figure 2 is an example of estimate specification data. In the calculation method M1 according to the embodiment, first, quotation specification data indicating the specifications of the item to be estimated is acquired (step S1). For example, as shown in FIG. 2, the quotation specification data 10A includes specific numerical values (specification values) of at least any of a plurality of specification items 11-1 to 11-n, requirements 12, etc. for the item to be estimated. The item is, for example, a product. The item may be a unit, module, or component included in the product. The item may be a unit group combining a plurality of units, a module group combining a plurality of modules, or a component group combining a plurality of components.

[0011] Generally, an item has various characteristics. Those characteristics are specified by the specification values of the specification items. At the order placement stage, as shown in FIG. 2, many specifications may be undetermined. Also, instead of specifying the specifications, ambiguous requirements 12 may be specified for the item.

[0012] After the acquisition of the quotation specification data, the requirements are converted into specific specification items (step S2). In the example shown in FIG. 2, as a requirement 12, "being robust" is required. Two specification items related to "being robust" are selected from the plurality of specification items 11-1 to 11-n. Thereby, the requirement 12 is converted into the two selected specification items (the first specification item and the second specification item). Or, three specification items may be selected from the plurality of specification items, and the requirement may be converted into three specification items (the first specification item, the second specification item, and the third specification item). As an example, the requirement 12 is converted into two specification items, namely, the "rotation speed" of the specification item 11-1 and the "size" of the specification item 11-4.

[0013] FIG. 3 is an example of case data. After the requirements are converted, several examples related to the item to be estimated are referenced. From each referenced example, the specification values ​​of the converted specification items are extracted (step S3). The examples include the amount, the specification values ​​of each specification item of the similar item, etc., for similar items similar to the item to be estimated. The examples are actual examples of similar items that have been produced or repaired in the past. Some of the multiple examples may be hypothetical examples prepared for the purpose of creating the model. For example, as shown in Figure 3, the example data 10B includes the specific specification values ​​of specification items 11-1 to 11-n, and the amount 13.

[0014] Figure 4 shows an example of a mapped distribution. The extracted specification values ​​are mapped onto a two-dimensional plane with the first and second specification items as the reference axes (step S4). If the requirements are converted into three specification items, the extracted specification values ​​are mapped onto a three-dimensional space with the first, second, and third specification items as the reference axes. For example, as shown in Figure 4, a distribution (scatter plot) 20 is obtained by mapping the specification value of "rotation speed" to the specification value of "magnitude" for each case.

[0015] Figure 5 is a schematic diagram illustrating an example of a method for extracting case examples. From the distribution created by mapping, a portion of multiple cases are extracted (step S5). Specifically, multiple cases are clustered based on the distribution. For example, as shown in Figure 5, a person sets dividing lines 21 and 22 on the distribution to cluster multiple cases. Alternatively, multiple cases may be pre-clustered for each specification item or combination thereof. Unsupervised learning machine learning can be used for clustering. For example, the k-means method is used. Dividing lines 21 and 22 can be set from the clustering results based on the specification values ​​of the first specification item, the clustering results based on the specification values ​​of the second specification item, or the clustering results based on a combination of the specification values ​​of the first and second specification items.

[0016] As an example, dividing lines 21 and 22 classify each case in distribution 20 into one of the regions 23-1 to 23-3. Region 23-1 is a region where the rotation speed is slow relative to the size. Region 23-2 is a region where the rotation speed is fast relative to the size. Region 23-3 is an intermediate region between regions 23-1 and 23-2. The smaller the item and the slower the rotation speed, the less wear occurs on each part of the item due to rotation. Therefore, the lifespan of the item or the period until repair is extended, and the item can be said to be "robust". In accordance with requirement 12, cases classified into region 23-1 are extracted from multiple cases. Here, some of the cases extracted from multiple cases are called "extracted cases". The number of dividing lines can be changed as appropriate. The number of dividing lines can be adjusted so that the number of extracted cases is appropriate and the estimation model described later can be created appropriately.

[0017] Apart from the extraction of case examples, related specification items are referenced for the item being estimated, selected from multiple specification items. Related specification items are those that have a relationship with the item. For example, specification items for which the customer has specified a value are selected as related specification items. Specification items that are considered to have a high degree of relevance to the item based on customer requirements may also be selected as related specification items.

[0018] In the example shown in Figure 2, a specification value is specified for specification item 11-n, "Capacity." Here, specification items for which a specification value is specified for the item being estimated are called "specified specification items." Specified specification items are an example of related specification items. The following section describes the case where there are specified specification items for which a specification value has been specified by the customer. There may be multiple specified specification items for the item being estimated. One or two specified specification items are selected from these specified specification items (step S6). In this case, it is preferable to select the specified specification items that have a greater impact on the price. If there are only one or two specified specification items, step S6 can be omitted.

[0019] Figure 6 is a schematic diagram illustrating the created estimation model. From each extracted example, the amount and the specification value of the selected specified specification item are extracted. Then, an estimation model showing the relationship between multiple amounts and multiple specification values ​​is created (Step S7). For example, as shown in Figure 6, multiple amounts and multiple specification values ​​(capacity) are mapped, and a distribution 30 is created. The estimation model is an approximation formula 31 created by multiple regression analysis. In the illustrated example, the approximation formula 31 is a linear equation showing the relationship between capacity and amount. The approximation formula 31 may also be a polynomial of degree two or higher. The base equation for the estimation model may be prepared in advance.

[0020] Alternatively, the estimation model may be created using machine learning. Machine learning can utilize regression methods such as stochastic gradient descent (SGD), LASSO, ElasticNet, Ridge, Support Vector Regression (SVR) using a Gaussian kernel, Linear SVR, and Ensemble.

[0021] The estimation model created through the above steps is then populated with the specification values ​​of the specified specifications for the item. By inputting this information into the created estimation model, the estimated price is calculated (step S8). For example, as shown in Figure 6, the price pr1 is calculated from the specification value sp1 of the specified specifications for the item and the approximation formula 31.

[0022] The calculated estimated amount is checked to see if it is reasonable (Step S9). For example, if the estimated amount is checked to be reasonable, it is presented to the customer. If the estimated amount is not reasonable, the creation method may be repeated under different conditions. For example, Step S5 may be repeated. At least some of the cases extracted in the newly executed Step S5 are different from at least some of the cases extracted in the previous Step S5. Alternatively, in Step S2, the requirements may be converted into different specification items. The "different specification items" are different from the specification items converted in the previously executed Step S2.

[0023] Figure 7 shows another example of a mapped distribution. Figure 8 is a schematic diagram showing another example of a case extraction method. Another example of the creation method according to the embodiment will be described. For example, the requirement is specified as "good cost performance of power generation." "Good cost performance" can be rephrased as "good efficiency in the desired output range." Therefore, in step S2, the requirement is converted into two specification items: specification item 11-2 "output" and specification item 11-3 "efficiency." As shown in Figure 7, the specification values ​​of each case are mapped using "output" and "efficiency" as reference axes, and a distribution 40 is created. As shown in Figure 8, dividing lines 41 and 42 are set for the distribution 40. Each case is classified into one of the regions 43-1 to 43-3.

[0024] The efficiency of power generation depends on the output range. In the example shown in Figure 8, region 43-1 is the region where the efficiency of power generation is low relative to the output. Region 43-2 is the region where the efficiency of power generation is high relative to the output. Region 43-3 is the intermediate region between regions 43-1 and 43-2. ​​The higher the efficiency of power generation relative to the output, the better the "cost performance." Based on the specified requirements, examples classified as region 43-2 are extracted from multiple examples.

[0025] As shown in Figures 4 and 7, "specification items" are details such as the specifications, performance, and functions of an item. In addition, the price of an item may be used as a "specification item." In this case, the specific value of the price is used as the "specification value." For example, if "being inexpensive" is specified as a requirement, the specification values ​​of multiple cases are mapped onto a two-dimensional plane or space with "price" and specification items other than price as the reference axes. A portion of the multiple cases are extracted from the distribution created by the mapping.

[0026] Subsequently, as in the example described above, the amount and specification values ​​of the specified specification items are extracted from the sample data, and an estimation model is created. The estimated amount is then calculated using the created estimation model.

[0027] (Advantages of the embodiment) Traditionally, the following method exists: First, an approximation function showing the relationship between specification values ​​and price is created using methods such as multiple regression and linear approximation, based on specification values ​​and price from past examples. Next, the price is calculated by inputting the specification values ​​of the item to be estimated into the created approximation function. With this method, there is no need to perform design work for estimation, and the price can be estimated in a short period of time by using the approximation function.

[0028] On the other hand, for custom-made indented products, the design is based on various concepts, such as the natural environment in which they will be installed and the required service life. Therefore, simply collecting past examples is insufficient, as the distribution of the population is wide-ranging, making it difficult to create an estimation model with the desired accuracy. In addition, customer requirements can sometimes be ambiguous, requiring time to confirm specifications or design the items to be estimated.

[0029] To address these issues, the creation method according to the embodiment first converts ambiguous requirements for the item to be estimated into first specification items and second specification items. By converting the requirements into specific specification items, it becomes possible to estimate the price while taking the requirements into account. Note that "ambiguous" refers to the specification or characteristics of an item being specified using strings other than numerical values. An example of an ambiguous requirement is when the specifications are described using linguistic data (especially adjectives). "Specific specifications are better than previous ones" is also an example of an ambiguous requirement.

[0030] Next, a distribution is created of the specification values ​​for the first specification item and the specification values ​​for the second specification item across multiple cases. Based on this distribution, some cases are extracted from the multiple cases. By using the transformed distribution of specification items, it is possible to extract cases that are highly relevant to the requirements. In addition, it is possible to prevent the distribution of the population from becoming too broad.

[0031] Next, the specified specifications for the item being estimated are referenced. The amounts in the selected examples and the specification values ​​of the specified specifications are referenced, and an estimation model showing their relationship is created. By using only the data from the selected examples and suppressing the spread of the population distribution, a more accurate estimation model is created.

[0032] The generated estimation model can calculate a more accurate estimated price based on the input of specification values ​​for the specified specifications of the item being estimated.

[0033] According to the creation method of the embodiment described above, ambiguous requirements are converted into specification items, and the estimated price is calculated using the specification values ​​of the specified specification items. Therefore, there is no need to design undetermined specification values ​​for the items to be estimated. Furthermore, since some cases are extracted from multiple cases using the specification items into which the requirements have been converted, the accuracy of the amount calculated by the estimation model can also be improved. According to the embodiment, an estimation model capable of calculating the estimated price with greater accuracy can be created.

[0034] Each step of the calculation method according to the embodiment may be performed by a person or by an electronic computing device (computer). Some of the steps may be performed by a person, and other parts of the steps may be performed by a computer.

[0035] (Computation system) Figure 9 is a schematic diagram showing a calculation system according to an embodiment. The calculation method described above can use, for example, the calculation system 100 shown in Figure 9. The calculation system 100 comprises a processing unit 110, a storage device 120, an input device 130, and an output device 140.

[0036] The processing unit 110 includes the functions of an acquisition unit 111, a conversion unit 112, an extraction unit 113, a selection unit 114, a model creation unit 115, a calculation unit 116, a verification unit 117, and an output unit 118. The storage device 120 stores data necessary for the creation method, data obtained by the creation method, etc. The input device 130 is used by the user to input data to the processing unit 110. The output device 140 outputs the data transmitted from the processing unit 110 in a way that the user can recognize.

[0037] The acquisition unit 111 executes step S1 of the calculation method M1. For example, the acquisition unit 111 acquires the estimate specification data 121 stored in the storage device 120. The estimate specification data 121 may also be entered by the user using the input device 130. The acquisition unit 111 accepts the estimate specification data 121 entered by the user.

[0038] The conversion unit 112 executes step S2. The conversion unit 112 converts the requirements contained in the estimate specification data 121 into specification items. For example, a rule database (DB) 122 is prepared in advance and stored in the storage device 120. The rule database 122 defines the correspondence between strings that may be included in the requirements and specification items. The conversion unit 112 converts the requirements in the estimate specification data 121 into specification items by referring to the rule database 122. Alternatively, a person may convert the requirements into specification items instead of the conversion unit 112.

[0039] The extraction unit 113 executes steps S3 to S5. Specifically, the extraction unit 113 extracts the specification values ​​of the converted specification items from the case database 123 stored in the storage device 120. The case database 123 includes data on actual examples of similar items that have been produced or repaired in the past. The case database 123 may also include data on hypothetical examples created based on actual examples. For example, hypothetical examples can be created by pre-calculating expected specification changes and the resulting changes in price based on actual examples. If the number of actual examples is insufficient, using hypothetical examples in addition to actual examples can improve the accuracy of the estimation model and allow for the calculation of a more appropriate estimated price. The extraction unit 113 maps the extracted specification values ​​and creates a distribution. Based on the distribution, the extraction unit 113 extracts some examples.

[0040] The selection unit 114 executes step S6. The selection unit 114 selects some of the specified specification items from a plurality of specified specification items. For example, a weight is set in advance for each specification item. The higher the weight, the greater the impact on the price. The selection unit 114 refers to the weight of each specified specification item and selects the specified specification item with the highest weight. Alternatively, a person may select the specified specification items on behalf of the selection unit 114.

[0041] The model creation unit 115 executes step S7. The model creation unit 115 creates an estimated model 124 using the amounts and specification values ​​of the extracted examples. The created estimated model 124 is stored in the storage device 120. The calculation unit 116 executes step S8. The calculation unit 116 calculates the estimated amount using the estimated model 124.

[0042] The verification unit 117 executes step S9. The output unit 118 verifies whether the calculated estimated amount is reasonable. For example, the verification unit 117 determines that the estimated amount is reasonable if it falls between the minimum and maximum values ​​of the extracted examples.

[0043] The output unit 118 outputs the calculated estimated amount. For example, if the output device 140 is a monitor, the output unit 118 displays the estimated amount on the monitor. Instead of the verification unit 117, a person may verify whether the estimated amount displayed on the output device 140 is reasonable. The output unit 118 may also save the estimated amount to a data sheet on a server or other device.

[0044] (User interface) Figures 10 and 11 are schematic diagrams illustrating a user interface. The processing unit 110 may display a user interface (UI) for inputting quotation specification data on the output device 140. For example, the processing unit 110 displays the UI 200 shown in Figure 10 on the output device 140. The UI 200 displays specification items 210 and input fields 220. Each cell in the column of specification items 210 contains specification items 211 to 217 related to the item to be quoted. Data is entered in the input fields 220 for each of the specification items 211 to 217.

[0045] The user uses the input device 130 to input data into the input field 220 while operating the pointer 230, etc. As shown in Figure 10, a pull-down menu 240 may be displayed in cells where specification values ​​can be entered. For example, by hovering the pointer 230 over the pull-down menu 240 and clicking the pull-down menu 240, a list of inputtable data is displayed, as shown in Figure 11. The user can then select the data to input from the list.

[0046] In the example shown in Figure 11, if data 241, which is a "numerical input," is selected, the user enters a specific specification value for that specification item. If data 242-244 is selected, past cases identified by one or more specification items 211-213 are referenced. The requirement is set to satisfy the selected conditions by comparing them with the specification values ​​of those past cases. If any of data 242-244 is entered for two or more specification items 214-217, the requirement is set to satisfy multiple selected conditions by comparing them with the specification values ​​of past cases. The requirement can be entered in any of specification items 211-217, or it can be entered in the cell for specification item 218. For example, in the example shown in Figure 2, "long" is entered in the cell for specification item 217, and "robust" is entered in the cell for specification item 218. In the example shown in Figure 7, "high" is entered in the cell for specification item 216, and "good cost performance of power generation" is entered in the cell for specification item 218.

[0047] Once the user has finished entering data into the input field 220, they click the icon 250. In response to the click of the icon 250, the processing unit 110 saves the entered data as the estimate specification data 121. In response to the click of the icon 250, the calculation method according to the embodiment may be automatically executed by the calculation system 100.

[0048] As shown in Figures 10 and 11, users can input specific specification values ​​for specification items where such values ​​are specified, through the UI200. If the customer has given vague requests, the user can input conditions or strings of specification values ​​corresponding to those requirements. Using the UI200, users can efficiently create quotation specification data for the items being quoted.

[0049] (Reliability) Each case may be assigned a confidence level. The confidence level is data indicating the credibility of each case. For example, a higher confidence level indicates that the case is useful for creating an estimation model. The confidence level is set according to the error between the estimated price and the actual price when goods were produced or repaired in the past. The smaller the error, the higher the confidence level. The confidence level may also be set based on the profit margin and manufacturing period of past cases. The confidence level may also be set based on two or more of the aforementioned error, profit margin, and manufacturing period. A low confidence level may be set for special cases that are not suitable for calculating estimated prices.

[0050] The assigned confidence level is used when creating the estimation model. For example, a threshold is set for the confidence level. Using the confidence level and threshold, a subset of cases may be extracted from multiple extracted cases.

[0051] Figure 12 is a schematic diagram illustrating a method for creating an estimation model using confidence levels. Figure 12 shows the distribution 50 between the specification values ​​and amounts of specified specification items for several sample cases. As an example, low confidence levels are set for cases 51a and 51b in Figure 12. Cases 51a and 51b are excluded when creating the estimation model. An estimation model (approximation formula) 52 is created based on the other cases excluding cases 51a and 51b.

[0052] By using confidence levels, it is possible to create estimation models that can calculate estimated costs with greater accuracy.

[0053] An estimation model may be created by weighting each data point based on its confidence level and then using machine learning modeling that reflects these weights.

[0054] Figure 13 is a schematic diagram illustrating how to create an estimation model using machine learning with confidence levels. For example, as shown in Figure 13, the estimation model includes a neural network 310. The neural network 310 includes an input layer 311, a hidden layer 312, and an output layer 313.

[0055] The training data 320 is used to train the estimation model. The training data 320 includes multiple cases 321a to 321n, labels 322a to 322n for each case, and confidence levels 323a to 323n for each case. Each of the cases 321a to 321n contains multiple specification values. Labels 322a to 322n are estimated amounts for each of the cases 321a to 321n. Confidence levels 323a to 323n are data indicating the credibility of each of the cases 321a to 321n. Confidence levels 323a to 323n are set according to the difference between the estimated amount and the actual amount for each case. A smaller difference indicates higher credibility for that case.

[0056] In training the estimation model, deep learning is performed using the training data 320. The neural network 310 is trained to estimate the estimated amount based on the input specification values ​​using the training data 320. The evaluation value, which indicates the learning accuracy, is the difference between the estimated amount and the amount of the label. The smaller the difference and the smaller the evaluation value, the more accurately the neural network 310 has learned. During training, the neural network 310 is trained to make the evaluation value smaller. In addition, confidence levels 323a to 323n are used as weights for each case during training. The larger the confidence levels 323a to 323n, the more strongly the neural network 310 learns based on that case.

[0057] Figure 14 is a schematic diagram showing the input data to the estimation model and the output data from the estimation model. As shown in Figure 14, the trained neural network 310 is input with the specifications 315 of the item to be estimated. The neural network 310 outputs output data 330 in response to the input of multiple specification values. The output data 330 includes the estimated amount (estimated price) 331 and the evaluation value 332. The evaluation value 332 is a self-assessment value indicating the likelihood of the estimated amount 331.

[0058] (modified version) The above describes specific examples of situations where ambiguous requirements exist for the items being estimated. If no ambiguous requirements exist, steps such as requirement transformation may be omitted.

[0059] Figure 15 is a flowchart showing a method for calculating an estimated price that includes a manufacturing method according to a modified embodiment. As shown in Figure 15, the calculation method M2 relating to the modified example further includes steps S11 to S16. Compared to calculation method M1, calculation method M2 further includes processing when there are no ambiguous requirements and processing when confidence levels are used.

[0060] First, a threshold for the degree of confidence is set (step S11). Then, the estimated specification data is acquired (step S1). After acquiring the estimated specification data, it is determined whether there are any ambiguous requirements in the specifications (step S12). If there are ambiguous requirements, steps S2 and onward are executed, similar to calculation method M1.

[0061] If there are no ambiguous requirements, the specification items or combinations of specification items defined in the classification master are referenced (step S13). The absence of ambiguous requirements means that all specification values ​​with a relatively large impact on the price are specified. The classification master shows the clustering results of the specification values ​​for each case for each specification item or combination thereof. Machine learning, as described above, can be used for clustering. The classification for each specification item or combination thereof is pre-stored as the classification master.

[0062] From the estimation specification data, specification values ​​related to specification items or combinations thereof defined in the class classification master are extracted (step S14). From the specification items or combinations thereof defined in the class classification master, classes whose confidence level exceeds the threshold are extracted (step S15).

[0063] In step S15, the classes to which the extracted specification values ​​belong are searched (step S16). If there are multiple matching classes, the class with the highest confidence level may be selected. After classification, in step S5, the cases that belong to that class are extracted from the multiple cases. After the cases are extracted, steps S6 and beyond are executed in the same manner as calculation method M1.

[0064] Figure 16 is a schematic diagram showing the configuration of a calculation system according to a modified embodiment. In the modified calculation system 100a, the processing unit 110a further includes the function of a classification unit 119. The classification unit 119 clusters the specification values ​​of each case registered in the case database 123 according to specification items or combinations thereof. The classification unit 119 stores the results as a class classification master 125 in the storage device 120a.

[0065] Figure 17 shows another example of quotation specification data. In the estimation specification data 60 shown in Figure 17, specific specification values ​​are assigned to each of the multiple specification items 61-1 to 61-n. The estimation specification data 60 is clear and does not contain any ambiguous requirements.

[0066] Figure 18 shows an example of a classification master. The class classification master 70 shown in Figure 18 includes a class ID 71, a range 72, a class name 73, and a confidence level 74. The class ID 71 is a string used to identify each class. The range 72 indicates the range of specification values ​​for each class. The class name 73 is the name of each class. The confidence level 74 indicates the confidence level assigned to each class. In the example shown in Figure 18, "specification item n" is classified into three classes 70-1 to 70-3. In addition, combinations of "specification item n" and "specification item n+1" are classified into three classes 70-4 to 70-6.

[0067] For example, if the threshold for confidence is set to "0.5" in step S11, classes 70-1 to 70-3 with a confidence of "0.8" are extracted from the class classification master 70. In the estimation specification data 60 shown in Figure 17, the specification value "70" is specified for specification item n. From classes 70-1 to 70-3, "item n_1" corresponding to the specification value "70" is extracted. Subsequently, cases that fall under the category of "item n_1" are extracted from multiple cases.

[0068] When no ambiguous requirements exist, the accuracy of the calculated cost estimate can be improved by extracting examples using predefined class classifications. Furthermore, the effort required to convert requirements into specification items can be eliminated.

[0069] The above describes an example where the object is an item. The invention according to the above embodiment is also applicable when the object is a building such as a power plant or office building, or real estate such as land. For example, the requirements for the object to be estimated are converted into multiple specification items, and a portion of several cases are extracted based on the specification values ​​of those specification items. Using the extracted cases, an estimation model is created that shows the relationship between the specification values ​​of related specification items and the amount related to the object. According to this estimation model, the estimated amount of the object can be estimated with greater accuracy.

[0070] Figure 19 is a schematic diagram representing the hardware configuration. As the processing unit 110, for example, the computer 90 shown in Figure 19 is used. The computer 90 includes a CPU 91, ROM 92, RAM 93, storage device 94, input interface 95, output interface 96, and communication interface 97.

[0071] ROM92 stores programs that control the operation of computer 90. ROM92 contains the programs necessary for computer 90 to perform each of the processes described above. RAM93 functions as a memory area where the programs stored in ROM92 are loaded.

[0072] The CPU 91 includes processing circuits. The CPU 91 uses the RAM 93 as work memory and executes programs stored in at least one of the ROM 92 or the storage device 94. During program execution, the CPU 91 controls each component via the system bus 98 and performs various processes.

[0073] The memory device 94 stores data necessary for program execution and data obtained through program execution.

[0074] The input interface (I / F) 95 connects the computer 90 to the input device 95a. The input I / F 95 is, for example, a serial bus interface such as USB. The CPU 91 can read various data from the input device 95a via the input I / F 95.

[0075] The output interface (I / F) 96 connects the computer 90 to the output device 96a. The output I / F 96 is a video output interface such as Digital Visual Interface (DVI) or High-Definition Multimedia Interface (HPMI®). The CPU 91 can transmit data to the output device 96a via the output I / F 96 and display an image on the output device 96a.

[0076] The communication interface (I / F) 97 allows the computer 90 to connect with a server 97a located outside the computer 90. The communication I / F 97 is, for example, a network card such as a LAN card. The CPU 91 can read various data from the server 97a via the communication I / F 97.

[0077] The storage device 94 includes one or more selected from Hard Disk Drives (HDDs) and Solid State Drives (SSDs). The input device 95a includes one or more selected from a mouse, keyboard, microphone (voice input), and touchpad. The output device 96a includes one or more selected from a monitor, projector, printer, and speaker. Devices that have the functions of both input device 95a and output device 96a, such as a touch panel, may also be used.

[0078] Each process performed by the processing unit 110 may be implemented by a single computer 90 or by the cooperation of multiple computers 90. The storage device 94 may be used as storage device 120 or 120a. The input device 95a and the output device 96a may be used as input device 130 and output device 140, respectively.

[0079] The processing of the various data described above may be recorded as a program that can be executed by a computer on a magnetic disk (flexible disk and hard disk, etc.), an optical disk (CD-ROM, CD-R, CD-RW, DVD-ROM, DVD±R, DVD±RW, etc.), a semiconductor memory, or another non-transitory computer-readable storage medium.

[0080] For example, information recorded on a recording medium can be read by a computer (or embedded system). The recording format (storage format) of the recording medium is arbitrary. For example, a computer reads a program from the recording medium and has the CPU execute the instructions written in the program based on this program. In a computer, program acquisition (or reading) may be performed via a network.

[0081] The inventions according to these embodiments may include the following technical solutions. (Plan 1) The steps involve converting the requirements for the property to be estimated into the first specification item and the second specification item, A step of extracting some of the aforementioned cases based on the distribution of the specification values ​​of the first specification item and the specification values ​​of the second specification item in multiple cases related to the aforementioned property, The steps include: creating an estimation model that shows the relationship between the amount in some of the aforementioned cases and the specification values ​​of the aforementioned related specification items, by referring to the relevant specification items that are related to the aforementioned property; How to create a prepared document. (Plan 2) In the extraction step, Using the first specification item and the second specification item as reference axes, the specification values ​​of the first specification item and the specification values ​​of the second specification item in the multiple examples are mapped to create the distribution. In the distribution described above, a dividing line is set to classify the multiple cases, Extract some of the cases classified by the dividing line. The creation method described in Proposal 1. (Plan 3) The creation method according to Proposal 2, wherein the dividing line is set using the cluster classification results based on the first specification of the plurality of cases and the cluster classification results based on the second specification of the plurality of cases. (Plan 4) A method for creating a document according to one of the proposals 1 to 3, wherein in the conversion step, the requirements are converted into the first specification and the second specification selected from three or more specifications commonly set for the multiple cases. (Plan 5) In the conversion step, the requirements are converted into the first specification item, the second specification item, and the third specification item, which are selected from four or more specification items that are commonly set for the multiple cases. A method for creating a sample, as described in any one of proposals 1 to 3, wherein in the extraction step, some of the above-mentioned examples are extracted based on the distribution of the first specification item, the second specification item, and the third specification item. (Plan 6) A method for creating a model, as described in one of the proposals 1 to 5, which involves creating an estimation model in the step described above that shows the relationship between the amount in some of the cases and the specification values ​​of multiple related specification items. (Plan 7) The aforementioned item is a custom-made product, or a unit, module, or component included in the custom-made product, and the method of creation is as described in any one of proposals 1 to 6. (Plan 8) The method of creation described in Proposal 2, wherein the amounts in the aforementioned multiple examples are used as the first specification item or the second specification item. (Plan 9) Each of the aforementioned cases is assigned a different level of confidence. A method for creating a sample according to any one of the proposals 1 to 8, wherein in the extraction step, in addition to the distribution, a subset of cases is extracted using the multiple confidence levels. (Plan 10) The aforementioned multiple cases Examples of similar properties that are similar to the aforementioned property and have been produced in the past, A hypothetical case created based on the aforementioned example, The creation method described in one of the proposals 1-9, including the one mentioned above. (Plan 11) If the aforementioned requirements do not exist for the aforementioned property, the conversion step is omitted, and some of the aforementioned cases are extracted based on the distribution of the relevant specification items in the aforementioned multiple cases, as described in one of the proposals 1 to 10. (Plan 12) The creation method described in one of the options 1-11, The estimation model includes the step of calculating the estimated price of the property by inputting the specification values ​​of the relevant specification items in the property, A calculation method that includes [a specific feature / feature]. (Plan 13) A processing device that executes one of the creation methods described in Proposal 1 to 11. (Plan 14) A program that causes a computer to execute one of the creation methods described in one of the options 1-11. (Plan 15) A storage medium containing the program described in Proposal 14.

[0082] According to the embodiments described above, a method for creating a model capable of calculating estimated amounts with greater accuracy is provided, along with a calculation method, a processing unit, a program, and a storage medium.

[0083] Although several embodiments of the present invention have been illustrated above, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents. Furthermore, the embodiments described above can be implemented in combination with each other. [Explanation of symbols]

[0084] 10A: Estimate specification data, 10B: Case data, 11: Specification item, 12: Requirements, 20: Distribution, 21,22: Dividing line, 23-1~23-3: Region, 30: Distribution, 31: Approximation formula, 40: Distribution, 41,42: Dividing line, 43-1~43-3: Region, 50: Distribution, 51a,51b: Case, 52: Estimation model, 60: Estimate specification data, 61: Specification item, 70: Class classification master, 71: Class ID, 72: Range, 73: Class name, 74: Confidence level, 90: Computer, 91: CPU, 92: ROM, 93: RAM, 94: Storage device, 95: Input interface, 95a: Input device, 96: Output interface, 96a: Output device, 97: Communication interface, 97a: Server, 98: System bus, 100,100a: Calculation system, 110,100a: Processing unit, 111: Acquisition unit, 112: Conversion unit, 113: Extraction unit, 114: Selection unit, 115: Model creation unit, 116: Calculation unit, 117: Verification unit, 118: Output unit, 119: Classification unit, 120,120a: Storage device, 121: Estimate specification data, 122: Rule database, 123: Case database, 124: Estimation model, 125: Class classification master, 130: Input device, 140: Output device, 210~218: Specification items, 220: Input field, 230: Pointer, 240: Pull-down menu, 241~244: Data, 250: Icon, 310: Neural network 311: Input layer, 312: Hidden layer, 313: Output layer, 315: Specifications, 320: Training data, 330: Output data, 331: Estimated amount, 332: Evaluation value, M1, M2: Calculation method

Claims

1. A method of creation performed by a computer, The steps include converting the requirements for the property to be estimated into first specification items and second specification items by referring to the correspondence between the strings included in the requirements and the specification items, Using the first specification item and the second specification item as reference axes, the specification values ​​of the first specification item and the specification values ​​of the second specification item in multiple cases related to the property are mapped to create a distribution, a dividing line is set in the distribution to classify the multiple cases, and some of the cases classified by the dividing line are extracted. The steps include: creating an estimation model that shows the relationship between the amount in some of the aforementioned cases and the specification values ​​of the aforementioned related specification items, by referring to the relevant specification items that are related to the aforementioned property; How to create a prepared document.

2. The creation method according to claim 1, wherein the dividing line is set using the cluster classification results based on the first specification item of the plurality of cases and the cluster classification results based on the second specification item of the plurality of cases.

3. The creation method according to claim 1 or 2, wherein in the conversion step, the requirements are converted into a first specification item and a second specification item selected from three or more specifications commonly set for the plurality of cases.

4. In the conversion step, the requirements are converted into the first specification item, the second specification item, and the third specification item, which are selected from four or more specification items that are commonly set for the multiple cases. The creation method according to claim 1 or 2, wherein in the extraction step, some of the cases are extracted based on the distribution of the first specification item, the second specification item, and the third specification item.

5. The creation method according to claim 1 or 2, wherein in the step of creating the above, an estimation model is created that shows the relationship between the amount in some of the above cases and the specification values ​​of a plurality of the above-mentioned related specification items.

6. The method for creating a document according to claim 1, wherein the amounts in the plurality of cases are used as the first specification item or the second specification item.

7. Each of the aforementioned multiple cases is assigned a confidence level, and these confidence levels are set based on at least one selected from the group consisting of the error between the estimated amount and the actual amount, the profit margin, and the manufacturing period. The method for creating a sample according to claim 1 or 2, wherein, in the extraction step, in addition to the distribution, a subset of cases is extracted by comparing the threshold set for the plurality of confidence levels with the plurality of confidence levels.

8. A creation method performed by a computer, The steps include converting the requirements for the property to be estimated into first specification items and second specification items by referring to the correspondence between the strings included in the requirements and the specification items, A step in which multiple confidence levels are assigned, each set based on at least one selected from the group consisting of the error between the estimated amount and the actual amount, the profit margin, and the manufacturing period, and a step in which multiple cases related to the said item are referred to, A step of extracting some of the above cases based on a comparison of the threshold set for the above multiple confidence levels with the above multiple confidence levels, and the distribution of the specification values ​​of the first specification item and the specification values ​​of the second specification item in the above multiple cases, The steps include: creating an estimation model that shows the relationship between the amount in some of the aforementioned cases and the specification values ​​of the aforementioned related specification items, by referring to the relevant specification items that are related to the aforementioned property; How to create a prepared document.

9. The aforementioned multiple cases Examples of similar properties that have been produced in the past and are similar to the aforementioned property, A hypothetical case created based on the aforementioned example, A method of production according to any one of claims 1, 2, or 8, including the following:

10. The creation method according to any one of claims 1, 2, or 8, wherein if the requirements for the property do not exist, the conversion step is omitted, and some of the cases are extracted based on the distribution of the relevant specification items in the multiple cases.

11. A method of preparation according to any one of claims 1, 2, or 8, The estimation model includes the step of calculating the estimated price of the property by inputting the specification values ​​of the relevant specification items in the property, A calculation method that includes [a specific feature / feature].

12. A processing apparatus that performs the manufacturing method according to any one of claims 1, 2, or 8.

13. A program that causes a computer to execute the method of creation described in any one of claims 1, 2, or 8.

14. A storage medium storing the program described in claim 13.

Images