Abnormality detection method and abnormality detection system for cost in construction integration

Abstract

To provide an abnormality detection method and an abnormality detection system capable of contributing to the correction work support of an inputted construction amount by performing the abnormality detection of the construction amount based on actual results and outputting the contents of abnormality.SOLUTION: Receiving an input of diagnosis target cost data including a plurality of variables for each construction type, extracting normal cost data including the same plurality of variables from past construction design document data for each construction type, creating at least one normal area including the normal cost data for each construction type, calculating an abnormality degree indicating how much the diagnosis target cost data deviates from the normal area for each construction type, determining that the diagnosis target cost data is abnormal based on the abnormality degree, and outputting a determination result and the abnormality degree, the normal area creation step creates a normal area for each construction type on the basis of a learning model obtained by unsupervised learning of normal cost data for each construction type.SELECTED DRAWING: Figure 11

Description

[Technical Field]

[0001] The present invention relates to an anomaly detection method and an anomaly detection system, and more particularly to an anomaly detection method and an anomaly detection system that can detect cost anomalies in construction cost estimation. [Background technology]

[0002] Computer-based construction cost estimation systems have been known for some time. Construction cost estimation refers to the process of calculating an estimated price by accumulating the cost of construction, which is the product of the unit prices of machinery, labor, and materials (hereinafter collectively referred to as "machinery, labor, and materials") required for each type of construction work and their quantities. Here, machinery costs are the costs required to use machinery required to perform the work, labor costs are the costs of labor required to perform the work, and material costs are the costs of materials required to perform the work.

[0003] The construction cost estimation system calculates the construction cost and the estimated price, and outputs a construction specification document that lists the calculated construction cost and estimated price.

[0004] Orders and contracts are made based on the estimated price listed in the construction design documents, and especially when the construction work is a public works project, bidding is based on the estimated price. For this reason, it is extremely important to check whether the estimated price listed in the construction design documents is appropriate. However, conventional construction cost estimation systems also use manually entered data on unit prices and quantities, which can lead to errors in the estimated price.

[0005] Therefore, a method has been proposed for determining whether there is an input error in the construction cost described in the construction specification document (see, for example, Patent Document 1). The method disclosed in Patent Document 1 calculates the ratio of the construction cost for each type of construction in the construction specification document and the deviation from the average construction cost of past construction specification documents, and detects an input error in the construction cost when the absolute value of the deviation is greater than a threshold value. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Laid-Open No. 2018-206298 Summary of the Invention [Problem to be solved by the invention]

[0007] However, the conventional method disclosed in Patent Document 1 is based on a simple judgment of whether the deviation range is within the upper and lower thresholds, and does not judge whether the construction cost has been input incorrectly based on actual results.

[0008] For example, as shown in Figure 12, for three types of construction work A to C, such as road earthworks, ground improvement work, and paving work, the construction costs (black circles) listed in past construction design documents are distributed across multiple areas, and it is assumed that the average construction costs of the past construction design documents will fall between the multiple areas.

[0009] In such a situation, the average value does not correspond to actual results and is not appropriate as a criterion for determining the construction cost of new construction specifications, which are subject to input error judgment.

[0010] Furthermore, conventional methods such as that disclosed in Patent Document 1 had the problem that they did not show the user how much or small the construction costs of each type of construction in the new construction design document might be in error.

[0011] The present invention has been made to solve these conventional problems, and aims to provide an anomaly detection method and anomaly detection system that can detect anomalies in construction costs based on actual results and output the details of the anomaly, thereby contributing to supporting the correction work of input construction costs. [Means for solving the problem]

[0012] In order to solve the above problems, the anomaly detection method according to the present invention includes a diagnosis target cost data input step of receiving, for each construction type, input of diagnosis target cost data including a plurality of variables; a normal cost data extraction step of extracting, for each construction type, normal cost data including the same plurality of variables from past construction specification data; a normal area creation step of creating, for each construction type, one or more normal areas that include the normal cost data extracted by the normal cost data extraction step; and an anomaly degree calculation step of calculating, for each construction type, an anomaly degree that indicates how much the diagnosis target cost data deviates from the one or more normal areas. a determination step of determining that the diagnosed cost data for the corresponding construction type is normal when the degree of abnormality is equal to or less than a threshold, and determining that the diagnosed cost data for the corresponding construction type is abnormal when the degree of abnormality is greater than the threshold; a determination result output step of outputting the determination result by the determination step; and an abnormality degree output step of outputting the degree of abnormality calculated by the abnormality degree calculation step, wherein the normal area creation step is configured to create one or more normal areas for each construction type based on a learning model obtained by unsupervised learning of the normal cost data for each construction type.

[0013] With this configuration, the anomaly detection method of the present invention can use unsupervised learning to create a normal region of construction costs that is in line with actual results, and can detect anomalies in construction costs even if the normal region is separated into multiple regions.In particular, the anomaly detection method of the present invention can identify the location of an anomaly (input error) in more detail by detecting anomalies using multiple variables for each construction type.

[0014] Furthermore, the anomaly detection method according to the present invention outputs the normal / abnormal judgment result and the degree of anomaly for each construction project, making it possible to identify the types of construction projects with a high number of abnormalities, thereby contributing to supporting the correction work of the input construction costs.

[0015] In addition, the abnormality detection method according to the present invention may be configured such that the plurality of variables consist of three variables, and the judgment result output step displays the one or more normal regions and the diagnosed cost data for each type of construction work on a three-dimensional graph with the three variables as axes.

[0016] With this configuration, the anomaly detection method of the present invention outputs the diagnosed cost data that has been determined to be normal or abnormal together with the normal area in a three-dimensional graph, thereby making it possible to confirm which costs for machinery, labor, and materials have what kind of abnormalities, thereby contributing to supporting the correction work of the input construction costs.

[0017] In addition, the abnormality detection method of the present invention may be configured so that the multiple variables consist of three variables, and the judgment result output step displays the one or more normal areas and the diagnosed cost data for each construction type on a two-dimensional graph with two of the three variables as axes.

[0018] With this configuration, the anomaly detection method of the present invention outputs the diagnosed cost data that has been determined to be normal or abnormal, along with the normal range, onto a two-dimensional graph with, for example, two of the machinery and labor materials on each axis, thereby clearly displaying the extent to which there is a possibility of an input error in each of the machinery and labor materials, and contributing to supporting the correction work of the input construction costs.

[0019] Furthermore, the anomaly detection method according to the present invention may be configured such that the anomaly degree calculation step further calculates how the anomaly degree of the diagnosis target cost data changes in response to a change in one of the plurality of variables, and the judgment result output step displays the change in the anomaly degree calculated by the anomaly degree calculation step and the threshold value of the anomaly degree on a one-dimensional graph with each of the variables and the anomaly degree as axes, respectively.

[0020] With this configuration, the anomaly detection method of the present invention outputs, on a one-dimensional graph together with a threshold value, how the degree of anomaly of diagnosed cost data that has been determined to be normal or abnormal changes in response to a change in, for example, one of the variables of machinery, labor, and equipment, thereby clearly displaying the degree of potential input error in each machinery, labor, and equipment, and contributing to supporting the correction work of the input construction costs.

[0021] and an output unit that outputs the determination result by the determination unit and the abnormality degree calculated by the abnormality degree calculation unit. The normal area creation unit is configured to create one or more normal areas for each construction type based on a learning model obtained by unsupervised learning of the normal cost data for each construction type. [Effects of the Invention]

[0022] The present invention provides an anomaly detection method and an anomaly detection system that can detect anomalies in construction costs based on actual results and output the details of the anomaly, thereby contributing to supporting the correction work of input construction costs. [Brief explanation of the drawings]

[0023] [Figure 1] 1 is a diagram illustrating a configuration of an anomaly detection system according to an embodiment of the present invention. [Figure 2]1. FIG. 4 is a diagram showing an example of a price display screen displayed on an output device included in the anomaly detection system of FIG. [Figure 3] 2 is a diagram illustrating an example of functional blocks realized in a computing device included in the anomaly detection system of FIG. 1. FIG. [Figure 4] 2 is a graph showing an example of a normal region created by a computing device included in the anomaly detection system of FIG. 1. [Figure 5] FIG. 10 is a diagram showing an example of the display of the price display screen when there is a possibility of an input error in the construction price. [Figure 6] 1. FIG. 4 is a diagram showing an example of an abnormality level display screen displayed on an output device included in the anomaly detection system of FIG. [Figure 7] 2 is a diagram showing an example of a three-dimensional graph screen displayed on an output device included in the anomaly detection system of FIG. 1. FIG. [Figure 8] FIG. 2 is a diagram (part 1) showing an example of a two-dimensional graph screen displayed on an output device included in the anomaly detection system of FIG. 1. [Figure 9] FIG. 2 is a diagram (part 2) showing a display example of a two-dimensional graph screen displayed on an output device included in the anomaly detection system of FIG. 1. [Figure 10] 1. FIG. 4 is a diagram showing an example of a one-dimensional graph screen displayed on an output device included in the anomaly detection system of FIG. [Figure 11] 1 is a flowchart illustrating processing of an anomaly detection method using an anomaly detection system according to an embodiment of the present invention. [Figure 12] 1 is a graph for explaining a conventional anomaly detection method. DETAILED DESCRIPTION OF THE INVENTION

[0024] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of an anomaly detection method and an anomaly detection system according to the present invention will be described below with reference to the accompanying drawings.

[0025] Fig. 1 is a diagram showing an example of the hardware configuration of an anomaly detection system 1 according to this embodiment. As shown in Fig. 1, the anomaly detection system 1 is composed of a control device such as a computer having an input device 10 as an input unit, a calculation device 11, an output device 12 as an output unit, a communication device 13, and a storage device 14.

[0026] The input device 10 , the arithmetic unit 11 , the output device 12 , the communication device 13 , and the storage device 14 are connected to one another via a bus 20 .

[0027] The bus 20 is also connected to a network 30 via a communication device 13. The network 30 is, for example, an internet line or a dedicated communication line.

[0028] The communication device 13 is connected to the construction estimation system 200 via the network 30, and transmits and receives various data between the construction estimation system 200 and the anomaly detection system 1. For example, the communication device 13 is configured to receive new construction specification data (hereinafter also referred to as "new construction specification data") 141 from the construction estimation system 200.

[0029] The storage device 14 includes a ROM (Read Only Memory), a RAM (Random Access Memory), an HDD (Hard Disk Drive), etc., and stores various programs executed by the arithmetic device 11 and various data described later.

[0030] The input device 10 is composed of, for example, a keyboard, a mouse, etc. The input device 10 is configured to accept input of new construction cost data for each construction type on a price display screen displayed on the output device 12.

[0031] The output device 12 is, for example, a display or a printer, and is configured to output various display screens such as a price display screen. For example, new construction specification data 141 may be stored in the storage device 14 from the construction estimation system 200 via the communication device 13, and the output device 12 may read the new construction specification data 141 from the storage device 14 and display it on the price display screen.

[0032] Hereinafter, the data on the construction costs for each type of construction displayed on the price display screen will also be referred to as "diagnosis target cost data."

[0033] 2 is a diagram showing an example of a price display screen 40 displayed on the output device 12. The price display screen 40 displays input diagnosis target cost data acquired from the construction cost estimation system 200.

[0034] The network 30 may be equivalent to the bus of the construction estimation system 200, in which case it operates in unison with the construction estimation of the construction estimation system 200. In this case, during the estimation of the construction estimation system 200, the input device 10 may input the estimation, and immediately after the input of the estimation, the output device 12 may display whether or not there is an abnormality.

[0035] As shown in Fig. 2, each piece of cost data to be diagnosed has data items such as "Construction No. (Construction Number)," "Construction Type," "Machinery and Material Type," "Quantity," "Unit Price," and "Construction Amount." For example, in the "Construction Type" column on the amount display screen 40, if the construction type is earthwork, items such as "Excavation," "Transportation of Earth and Sand," "Leveling," "Road Body Embankment," and "Roadbed Embankment" are displayed.

[0036] 2, the diagnosis target cost data includes three variables: machine cost, labor cost, and material cost. Note that the number of variables included in the diagnosis target cost data is not limited to three, and may be any number equal to or greater than two.

[0037] Although not shown in Fig. 2, a screen on which the user can input "quantity" and "unit price" for specific items at a lower level than "type of construction" on the price display screen 40 in Fig. 2 may be displayed separately on the output device 12. In this case, the preprocessing unit 110, which will be described later, may be configured to link the contents of the user's input for the items at the lower level to machinery and materials, and display them on the price display screen 40.

[0038] The arithmetic device 11 is, for example, a central processing unit (CPU) or a graphics processing unit (GPU), and realizes various functions of the anomaly detection system 1 by executing programs stored in the storage device 14.

[0039] 3 is a diagram showing an example of functional blocks realized in the arithmetic device 11. When the arithmetic device 11 executes the abnormality detection program 140 stored in the storage device 14, the arithmetic device 11 realizes a preprocessing unit 110, a learning model creation unit 111, a normal region creation unit 112, an abnormality degree calculation unit 113, and a determination unit 114.

[0040] The pre-processing unit 110 links the new construction design document data 141 displayed on the price display screen 40 to classifications such as machine costs, labor costs, and material costs for each type of construction work, and generates cost data to be diagnosed.

[0041] Similarly, the preprocessing unit 110 links the diagnosed cost data displayed on the price display screen 40 to classifications such as machine costs, labor costs, and material costs for each construction type, and extracts normal cost data 142 from the past construction specification data. Here, the normal cost data 142 is data on construction costs that constitute past construction specification data in which all of the listed construction costs can be considered normal. In other words, each normal cost data 142 contains the same multiple variables as the diagnosed cost data.

[0042] As the past construction specification data, for example, data created in the past by the construction cost estimation system 200 can be used. The normal cost data 142 extracted by the preprocessing unit 110 is stored in the storage device 14.

[0043] Furthermore, the preprocessing unit 110 converts the extracted normal cost data 142 into a format that can be used for learning by the learning model creation unit 111. For example, the preprocessing unit 110 performs scaling processing, such as standardization of construction costs, on the extracted normal cost data 142 as necessary. Standardization is processing that shapes the normal cost data 142 so that the distribution of each variable that makes up the normal cost data 142 has an average of 0 and a standard deviation of 1.

[0044] The learning model creation unit 111 is configured to create a learning model 143 by performing unsupervised learning of the normal cost data 142 for each construction type using a learning algorithm for unsupervised learning, such as OC-SVM (One Class Support Vector Machine). The learning model 143 created by the learning model creation unit 111 is stored in the storage device 14.

[0045] The normal area creation unit 112 is configured to create one or more normal areas containing normal cost data 142 for each construction type based on the learning model 143 for each construction type created by the learning model creation unit 111.

[0046] Figure 4 is a graph showing an example of a normal region created by the normal region creation unit 112 for a certain construction type. In the graph of Figure 4, each black circle represents normal cost data 142, and the shaded region represents the normal region. When the normal cost data 142 is expressed using three variables, namely, machine cost, labor cost, and material cost, the normal region is expressed as a three-dimensional region in a three-dimensional space with these variables as its axes.

[0047] The abnormality degree calculation unit 113 uses the learning model 143 created by the learning model creation unit 111 to calculate the abnormality degree indicating how much the diagnosed cost data for each construction type displayed on the price display screen 40 deviates from one or more normal ranges for the same construction type.

[0048] The boundary line of the normal region represents the threshold value of the degree of abnormality (e.g., 0). For example, the further the diagnostic target cost data is from the boundary surface of the normal region, the larger the positive value of the degree of abnormality. On the other hand, the further the diagnostic target cost data is within the boundary surface of the normal region, for example, the closer it is to the inside of the normal region, the larger the negative value of the degree of abnormality.

[0049] When the degree of abnormality calculated by the abnormality degree calculation unit 113 for the diagnosed cost data displayed on the amount display screen 40 is below a threshold, the judgment unit 114 judges that the diagnosed cost data for the corresponding construction type is normal, and when the degree of abnormality calculated by the abnormality degree calculation unit 113 is greater than the threshold, the judgment unit 114 judges that the diagnosed cost data for the corresponding construction type is abnormal.

[0050] Here, when the threshold value is 0, the determination unit 114 determines whether the diagnosis target cost data is normal or abnormal based on whether the diagnosis target cost data is included in one or more normal regions for a specific construction type stored in the storage device 14. Note that the value of the threshold value is not limited to 0, and can be set to any value by the user using the input device 10.

[0051] The determination by the determination unit 114 is made when the user presses the decision button 41 on the price display screen 40 using the input device 10. When the determination unit 114 determines that all of the diagnosis target cost data currently displayed on the price display screen 40 is normal, the diagnosis target cost data may be transmitted from the communication device 13 to the construction cost estimation system 200.

[0052] On the other hand, when the determination unit 114 determines that there is an abnormality, the output device 12 indicates which "construction No." construction has a possible input error on the price display screen 40. In other words, the output device 12 is configured to output the determination result by the determination unit 114 on the price display screen 40.

[0053] The determination unit 114 may always determine whether the diagnosis target expense data is normal or abnormal, regardless of whether the decision button 41 is pressed or not.

[0054] 5 shows an example in which the output device 12 displays on the amount display screen 40 a message indicating that there may be an input error in the "construction cost" input for a construction project with a "construction number" of 1001. For example, the output device 12 highlights the field for "construction number" 1001, which has been determined by the determination unit 114 to be a possible input error, and displays a link 42 for guiding the user to a detailed information display screen that displays detailed information about the input error. Note that the output device 12 may output a message indicating there may be an input error by other methods, such as a pop-up display, rather than by highlighting the field.

[0055] Link 42 displays a message informing of the possibility of an input error, such as "Warning: There may be an input error in construction No. 1001."

[0056] 6 to 10 show examples of detailed information display screens that are output by the output device 12 when the user presses a link 42 on the amount display screen 40 using the input device 10. FIG.

[0057] As shown in Fig. 6, the abnormality degree display screen 50, which serves as a detailed information display screen, displays bar graphs showing machine costs, labor costs, material costs, and the abnormality degree, with the horizontal axis representing "construction number." Fig. 6 shows an example in which the data of each bar graph is arranged in descending order of "construction number" with the highest degree of abnormality on the abnormality degree display screen 50, but the display format is not limited to this, and a display format in which the data of each bar graph is arranged in descending order of "construction number" may also be used.

[0058] That is, the abnormality degree display screen 50 visualizes the abnormality degree for each construction project, and allows the user to easily understand which "construction No." construction project has the greatest abnormality.

[0059] The abnormality degree display screen 50 includes a plurality of screen display buttons 51 to 54 for displaying various detailed information display screens showing detailed information on the diagnosis target cost data that the determining unit 114 has determined to be abnormal.

[0060] For example, when a user uses the input device 10 to press the three-dimensional graph button 52 on a detailed information display screen such as the abnormality level display screen 50, the output device 12 displays a three-dimensional graph screen 60 as the detailed information display screen, as shown in Figure 7.

[0061] The three-dimensional graph screen 60 displays, in a three-dimensional graph, one or more three-dimensional normal regions for the construction type of diagnosed cost data that has been determined to be abnormal by the determination unit 114, and the position in three-dimensional space of the diagnosed cost data that has been determined to be normal or abnormal by the determination unit 114. The three-dimensional graph displayed on the three-dimensional graph screen 60 has three variables, namely, machine cost, labor cost, and material cost, as axes for each construction type.

[0062] If there are multiple types of construction work for which the diagnosis target cost data has been determined to be abnormal by the determination unit 114, buttons or the like are displayed on the 3D graph screen 60 to switch between the normal area and the construction type for the diagnosis target cost data.

[0063] The three-dimensional graph screen 60 may be configured to display the diagnosed cost data of the construction type to be displayed in a differentiated manner according to the determination result by the determination unit 114. For example, on the three-dimensional graph screen 60, of the diagnosed cost data displayed on the amount display screen 40, data that is determined to be normal by the determination unit 114 may be displayed with a circle, and data that is determined to be abnormal by the determination unit 114 may be displayed with an x.

[0064] When a user uses the input device 10 to press the two-dimensional graph button 53 on a detailed information display screen such as the abnormality level display screen 50, the output device 12 displays a two-dimensional graph screen 70 as the detailed information display screen, as shown in Figures 8 and 9.

[0065] The two-dimensional graph screen 70 displays, in a two-dimensional graph, one or more two-dimensional normal areas for the construction type of diagnosed cost data that has been determined to be abnormal by the determination unit 114, and the position in two-dimensional space of the diagnosed cost data that has been determined to be abnormal by the determination unit 114.

[0066] If there are multiple types of construction work for which the diagnosis target cost data has been determined to be abnormal by the determination unit 114, buttons or the like are displayed on the two-dimensional graph screen 70 to switch between the normal area and the construction type for the diagnosis target cost data.

[0067] FIG. 8 shows an example of a display on the two-dimensional graph screen 70 for diagnostic target cost data in which only material costs are abnormal and machine costs and labor costs are normal.

[0068] The graph G1 on the left is a graph with the horizontal axis representing machine costs a1 and the vertical axis representing labor costs a2, out of the three variables, and shows a slice plane parallel to the a1a2 plane containing abnormal diagnostic target cost data in a three-dimensional graph such as that shown in Figure 4.

[0069] In graph G1, the machine costs and labor costs of the cost data to be diagnosed are plotted with an X mark as abnormal input data. There is no normal area for the material cost values ​​of the cost data to be diagnosed. Graph G1 shows that even if the machine costs and labor costs are changed while the material costs are fixed, the cost data to be diagnosed will not become normal.

[0070] The central graph G2 is a graph with labor cost a2 on the horizontal axis and material cost a3 on the vertical axis, out of the three variables, and shows a slice plane parallel to the a2a3 plane containing abnormal diagnosis target cost data in a three-dimensional graph such as that shown in Figure 4.

[0071] In graph G2, the material cost of the diagnostic object cost data is plotted as abnormal input data with an X. The shaded area surrounded by a solid line indicates the normal area on the slice surface.

[0072] Graph G2 also displays the message "Material costs are abnormal" and an arrow indicating how much the material costs need to be changed to normalize them. Graph G2 shows that if the material costs are reduced along the dashed arrow, the cost data subject to diagnosis will return to normal.

[0073] Graph G3 on the right is a graph with material cost a3 on the horizontal axis and machine cost a1 on the vertical axis, out of the three variables, and shows a slice plane parallel to the a3a1 plane containing abnormal diagnosis target cost data in a three-dimensional graph such as that shown in Figure 4.

[0074] In graph G3, the material cost of the diagnostic object cost data is plotted as abnormal input data with an X. The shaded area surrounded by a solid line indicates the normal area on the slice surface.

[0075] Graph G3 also displays the message "Material costs are abnormal" and an arrow indicating how much the material costs need to be changed to normalize them. Graph G3 shows that if the material costs are reduced along the dashed arrow, the cost data to be diagnosed will return to normal.

[0076] FIG. 9 shows an example of a display on the two-dimensional graph screen 70 for diagnosis target cost data in which labor costs and material costs are abnormal and only machine costs are normal.

[0077] In graph G4 on the left, labor costs in the cost data to be diagnosed are plotted as abnormal input data with an X. This graph G4 shows that even if machine costs and labor costs are changed while material costs are fixed, the cost data to be diagnosed will not become normal.

[0078] In the central graph G5, the labor costs and material costs of the diagnosis target cost data are plotted as abnormal input data with X marks. The shaded area surrounded by a solid line indicates the normal area on the slice surface.

[0079] Graph G5 also displays the message "Labor costs and material costs are abnormal" and arrows indicating how much the labor costs and material costs need to be changed to return them to normal. Graph G5 shows that if the labor cost value is increased along the dashed arrow and the material cost value is decreased along the dashed arrow, the cost data to be diagnosed will return to normal.

[0080] In graph G6 on the right, the material costs of the cost data to be diagnosed are plotted as abnormal input data with an X. This graph G6 shows that even if the material costs and machine costs are changed while the labor costs are fixed, the cost data to be diagnosed will not become normal.

[0081] The abnormality degree calculation unit 113 is further configured to calculate how the abnormality degree of the diagnosis target expense data that has been determined to be abnormal by the determination unit 114 changes in accordance with a change in one of the multiple variables.

[0082] When a user uses the input device 10 to press the one-dimensional graph button 54 on a detailed information display screen such as the abnormality level display screen 50, the output device 12 displays a one-dimensional graph screen 80 as the detailed information display screen, as shown in Figure 10.

[0083] The one-dimensional graph screen 80 is a screen that displays the change in the degree of abnormality calculated by the abnormality degree calculation unit 113 and the threshold value of the degree of abnormality in a one-dimensional graph for each variable, with the variable and the degree of abnormality as axes, respectively.

[0084] If there are multiple types of construction work for which the diagnosis target cost data has been determined to be abnormal by the determining unit 114, the one-dimensional graph screen 80 displays a button or the like for switching the type of construction work for which the abnormality level is to be displayed.

[0085] FIG. 10 shows an example of a one-dimensional graph screen 80 displaying the degree of abnormality of the diagnosis target cost data in which the labor cost and material cost are normal and only the machine cost is abnormal.

[0086] Graph G7 on the left shows the degree of abnormality when the labor cost and material cost of the cost data to be diagnosed are fixed and only the machine cost is changed. In graph G7, the machine cost of the cost data to be diagnosed is shown by a vertical line, and the threshold of the degree of abnormality is shown by a horizontal line.

[0087] Graph G7 displays an arrow indicating how much the machine cost needs to be changed to normalize the abnormality. Graph G7 shows that if the machine cost value of the diagnostic target cost data is increased along the dashed arrow, the diagnostic target cost data will become normal.

[0088] Graph G8 in the center is a graph showing the degree of abnormality when the machine cost and material cost of the diagnosis target cost data are fixed and only the labor cost is changed. In graph G8, the labor cost of the diagnosis target cost data is shown as a vertical line, and the threshold of the degree of abnormality is shown as a horizontal line. Graph G8 shows that the diagnosis target cost data does not become normal even if only the labor cost of the diagnosis target cost data is changed.

[0089] Graph G9 on the right shows the degree of abnormality when the machine cost and labor cost of the diagnosis target cost data are fixed and only the material cost is changed. In graph G9, the material cost of the diagnosis target cost data is shown as a vertical line, and the threshold of the degree of abnormality is shown as a horizontal line. Graph G9 shows that the diagnosis target cost data does not become normal even if only the material cost of the diagnosis target cost data is changed.

[0090] In this way, the one-dimensional graph screen 80 shows which of the equipment and materials may have an input error by plotting thresholds that indicate the boundary between the normal range and abnormal range of each variable in the cost data to be diagnosed, and the degree of abnormality of the cost data to be diagnosed.

[0091] When the user presses the abnormality degree button 51 using the input device 10 on a detailed information display screen other than the abnormality degree display screen 50, the output device 12 displays the abnormality degree display screen 50 again.

[0092] Furthermore, when the user presses the close button 55 on each detailed information display screen using the input device 10, the output device 12 closes the currently displayed detailed information display screen and displays the amount display screen 40 again.

[0093] In the above description, each detailed information display screen is configured as a separate screen, but the present invention is not limited to this, and for example, the contents of each detailed information display screen may be displayed together on a single screen. Also, for example, the amount display screen 40 and the contents of each detailed information display screen may be displayed together on a single screen.

[0094] An example of the processing of an anomaly detection method using the anomaly detection system 1 will be described below with reference to the flowchart in Fig. 11. Note that descriptions that overlap with the description of the configuration of the anomaly detection system 1 described above will be omitted as appropriate.

[0095] First, the input device 10 accepts input of diagnosis target cost data including a plurality of variables for each construction type on the price display screen 40 (step S1). Step S1 is a diagnosis target cost data input step.

[0096] Next, the input device 10 accepts pressing of the decision button 41 on the amount display screen 40 (step S2).

[0097] Next, the preprocessing unit 110 extracts, from the past construction specification data, normal cost data 142 for each construction type, which includes multiple variables identical to the multiple variables included in the diagnosed cost data displayed on the price display screen 40 (step S3). Step S3 is a normal cost data extraction step.

[0098] Next, the learning model creation unit 111 performs unsupervised learning on the normal cost data 142 extracted in step S3 for each construction type to create a learning model 143 (step S4).

[0099] Next, the normal area creation unit 112 creates one or more normal areas containing the normal cost data 142 for each construction type based on the learning model 143 created by the learning model creation unit 111 (step S5). Step S5 is a normal area creation step.

[0100] Next, the abnormality degree calculation unit 113 calculates the degree of abnormality indicating how much the diagnosed cost data for a specific construction type deviates from one or more normal regions for the specific construction type (step S6). In step S6, the abnormality degree calculation unit 113 calculates the degree of abnormality for the diagnosed cost data for all construction types displayed on the price display screen 40. Step S6 is an abnormality degree calculation step.

[0101] Next, the determination unit 114 determines whether the degree of abnormality calculated by the abnormality degree calculation unit 113 is greater than a threshold value (step S7). If the degree of abnormality is greater than the threshold value (step S7: YES), the process of step S8 is executed. If the degree of abnormality is equal to or less than the threshold value (step S7: NO), the series of processes ends.

[0102] In step S7, the determination unit 114 determines the magnitude of the degree of abnormality for the diagnosis target cost data of all construction types displayed on the price display screen 40. Step S7 is a determination step.

[0103] In step S8, the output device 12 displays on the price display screen 40 which "Construction No." construction work for which there is a possibility of an input error in the construction cost, as shown in Fig. 5. Step S8 is a determination result output step for outputting the determination result from the determination step.

[0104] Next, the input device 10 accepts pressing of the link 42 on the amount display screen 40 (step S9).

[0105] Next, the output device 12 outputs the degree of abnormality calculated by the abnormality degree calculation unit 113 for each "construction No." on the abnormality degree display screen 50 as shown in Fig. 6 (step S10). Step S10 is an abnormality degree output step.

[0106] Next, when any of the screen display buttons 51 to 54 is pressed (step S11: YES), the process of step S12 is executed.

[0107] In step S12, the output device 12 displays the abnormality level display screen 50, the three-dimensional graph screen 60, the two-dimensional graph screen 70, or the one-dimensional graph screen 80, depending on the screen display button pressed in step S11. Step S12 is a determination result output step.

[0108] If none of the screen display buttons 51 to 54 is pressed (step S11: NO) and the close button 55 is pressed (step S13: YES), the process of step S14 is executed.

[0109] In step S14, the output device 12 ends the display of the screen for which the close button 55 was pressed, among the abnormality degree display screen 50, the three-dimensional graph screen 60, the two-dimensional graph screen 70, and the one-dimensional graph screen 80. This ends the series of processes.

[0110] As described above, the anomaly detection method according to this embodiment creates one or more normal areas for each construction type based on a learning model that performs unsupervised learning on past construction design document data, and calculates the degree of anomaly for each construction type, indicating the degree to which the diagnosed cost data, including newly entered construction costs, deviates from one or more normal areas.

[0111] With this configuration, the anomaly detection method according to this embodiment can use unsupervised learning to create a normal region of construction costs that is in line with actual results, and can detect anomalies in construction costs even if the normal region is separated into multiple regions. In particular, the anomaly detection method according to this embodiment can identify the location of an anomaly (input error) in more detail by detecting anomalies using three-dimensional variables of machinery, labor, and materials (machine costs, labor costs, and material costs) for each construction type.

[0112] Furthermore, the anomaly detection method according to this embodiment outputs the normal / abnormal judgment result and a graph showing the degree of anomaly for each construction project, making it possible to identify construction projects with a high number of abnormalities, thereby contributing to supporting the correction of input construction costs.

[0113] In addition, the abnormality detection method according to this embodiment outputs the diagnosed cost data that has been determined to be normal or abnormal together with the normal region onto a three-dimensional graph, thereby enabling a clear display of which costs for machinery and materials have abnormalities and what kind of abnormalities there are, thereby contributing to supporting the correction work of the input construction costs.

[0114] In addition, the abnormality detection method according to this embodiment outputs the diagnosed cost data judged to be normal or abnormal, along with the normal range, onto a two-dimensional graph with two of the machinery and labor items as axes, thereby clearly displaying the degree of potential input error for each of the machinery and labor items, thereby contributing to supporting the correction work of the input construction costs.

[0115] Furthermore, the anomaly detection method according to this embodiment outputs, on a one-dimensional graph together with a threshold value, how the degree of anomaly of diagnosed cost data that has been determined to be normal or abnormal changes in response to changes in one of the variables of the machinery and labor, thereby clearly displaying the extent to which there is a possibility of an input error in which machinery and labor, thereby contributing to supporting the correction work of the input construction amount. [Explanation of symbols]

[0116] 1. Anomaly detection system 10 Input Devices 11 Arithmetic unit 12 Output Devices 13. Communications equipment 14 Storage device 20 Bus 30 Network 40 Amount display screen 42 Links 50 Abnormality degree display screen 60 3D graph screen 70 2D graph screen 80 1D graph screen 110 Pretreatment section 111 Learning Model Creation Department 112 Normal Region Creation Section 113 Abnormality calculation unit 114 Judgment section 140 Anomaly Detection Program 141 New Construction Design Data 142 Normal Cost Data 143 Learning Model 200 Construction Cost Estimation System

Claims

1. a diagnosis target cost data input step for receiving input of diagnosis target cost data including a plurality of variables for each construction type; a normal cost data extraction step of extracting normal cost data including the same plurality of variables for each construction type from past construction specification data; a normal area creation step of creating one or more normal areas for each construction type, the normal area including the normal cost data extracted by the normal cost data extraction step; an abnormality degree calculation step of calculating an abnormality degree indicating how much the diagnosis target cost data deviates from the one or more normal regions for each of the construction types; a determining step of determining that the diagnostic target cost data for the corresponding construction type is normal when the degree of abnormality is equal to or less than a threshold value, and determining that the diagnostic target cost data for the corresponding construction type is abnormal when the degree of abnormality is greater than the threshold value; a determination result output step for outputting a determination result obtained by the determination step; an abnormality degree output step of outputting the abnormality degree calculated by the abnormality degree calculation step, The anomaly detection method is characterized in that the normal area creation step creates one or more normal areas for each construction type based on a learning model that has undergone unsupervised learning of the normal cost data for each construction type.

2. the plurality of variables consists of three variables, The abnormality detection method according to claim 1, characterized in that the judgment result output step displays the one or more normal areas and the diagnosis target cost data for each construction type on a three-dimensional graph with the three variables as axes.

3. the plurality of variables consists of three variables, The abnormality detection method according to claim 1, characterized in that the judgment result output step displays the one or more normal areas and the diagnosis target cost data for each construction type on a two-dimensional graph with two of the three variables as axes.

4. The abnormality degree calculation step further calculates how the abnormality degree of the diagnostic object cost data changes in response to a change in one variable among the plurality of variables, 2. The anomaly detection method according to claim 1, wherein the judgment result output step displays the change in the degree of anomaly calculated in the degree of anomaly calculation step and the threshold value of the degree of anomaly on a one-dimensional graph with each of the variables and the degree of anomaly as axes, respectively.

5. an input unit that accepts input of diagnosis target cost data including a plurality of variables for each construction type; a pre-processing unit that extracts normal cost data including the same plurality of variables for each construction type from past construction specification data; a normal area creation unit that creates one or more normal areas containing the normal cost data for each of the construction types; an abnormality degree calculation unit that calculates an abnormality degree indicating how much the diagnosis target cost data deviates from the one or more normal regions for each of the construction types; a determination unit that determines that the diagnostic target cost data for the corresponding construction type is normal when the degree of abnormality is equal to or less than a threshold, and determines that the diagnostic target cost data for the corresponding construction type is abnormal when the degree of abnormality is greater than the threshold; an output unit that outputs the determination result by the determination unit and the abnormality degree calculated by the abnormality degree calculation unit, The normal area creation unit creates the one or more normal areas for each construction type based on a learning model that has undergone unsupervised learning of the normal cost data for each construction type.

Images