Detection program, method, and device

Abstract

This detection device extracts, out of a plurality of data containing one or more explanatory variables and an objective variable, specific conditions that have a predetermined correlation with the objective variable from among conditions that are combinations of the explanatory variables, and detects, from among a plurality of subsets of data divided into a first subgroup and a second subgroup on the basis of the specific conditions, subsets in which the sign of a regression coefficient between the objective variable and the explanatory variables changes between the first subgroup and the second subgroup as candidates for a volcano plot.

Description

Detection program, method, and device

[0001] The disclosed technology relates to a detection program, a detection method, and a detection device.

[0002] An input attribute condition determination device has been proposed to improve the efficiency of data analysis. This device calculates a first frequency and a second frequency for each of all numerical values ​​of an input attribute for an analysis data group classified into a first data group and a second data group based on the value of an output attribute. The first frequency and the second frequency are the proportions of data in the first data group and the second data group whose input attribute is equal to or less than the corresponding numerical value. This device also calculates a difference between the first frequency and the second frequency for each of all numerical values ​​of the input attribute. This device also determines a threshold value for the input attribute that maximizes the difference value among the numerical values ​​of the input attribute, and determines at least one threshold value corresponding to at least one input attribute. This device then determines an input attribute condition for dividing the analysis data group into two groups, the first data group and the second data group, based on the threshold value.

[0003] Furthermore, a data analysis device has been proposed that can efficiently extract significant features, statistically significant differences, and correlations between data. This device sorts records of data to be analyzed, each of which has an ordinal number serving as an explanatory variable and a corresponding objective variable, in ascending or descending order of the objective variable, and assigns new ordinal numbers to the sorted records. Based on regression tree analysis, the device then divides the sorted records into two split groups using a specific number selected from the assigned ordinal numbers as the boundary, and outputs the results.

[0004] JP 2004-252972 A JP 2010-153588 A

[0005] For example, in the field of materials, when data on the physical properties of a material are plotted in a region where a certain explanatory variable is on the X-axis and a target variable is on the Y-axis, the graph may have a mountain or valley shape. Such a graph is called a volcano plot, and knowing the peaks of the plot is important for understanding the physical properties.

[0006] However, to obtain an appropriate volcano plot, determining which explanatory variable to set on the X-axis among the various explanatory variables contained in the data relies heavily on the knowledge and experience of experts. It can also be difficult to grasp the trend of peaks or valleys from a scatter plot of data plotted on the data. It may be possible to detect a volcano plot more easily by setting conditions for the plotted data in addition to the explanatory variables set on the X-axis. However, when the number of explanatory variables is large, the number of possible conditions increases exponentially. Therefore, when such conditions are set manually, there is a risk that conditions may be overlooked, making comprehensive verification difficult.

[0007] In one aspect, the disclosed technique aims to detect volcano plots under various conditions without relying on the knowledge and experience of an expert.

[0008] In one aspect, the disclosed technology extracts, from a plurality of data sets including one or more explanatory variables and a dependent variable, specific conditions that have a predetermined correlation with the dependent variable from among conditions that are combinations of the explanatory variables. The disclosed technology also extracts subsets of the plurality of data sets that are divided into a first subgroup and a second subgroup based on the specific conditions. The disclosed technology then detects the subsets in which the positive and negative signs of the regression coefficients between the dependent variable and the explanatory variables change between the first subgroup and the second subgroup as candidates for a volcano plot.

[0009] One aspect is that it has the effect of making it possible to detect volcano plots under various conditions without relying on the knowledge and experience of an expert.

[0010] FIG. 1 is a diagram for explaining a volcano plot. FIG. 2 is a diagram showing a schematic configuration of a detection system and functional blocks of a detection device. FIG. 3 is a diagram for explaining extraction of specific conditions. FIG. 4 is a diagram for explaining extraction of specific conditions. FIG. 5 is a diagram for explaining detection of volcano plot candidates. FIG. 6 is a diagram showing an example of a detection result DB. FIG. 7 is a diagram showing a schematic configuration of a computer functioning as a detection device. FIG. 8 is a flowchart showing an example of detection processing. FIG. 9 is a diagram for explaining an application example. FIG. 10 is a diagram for explaining detection of a range of vertex candidates. FIG. 11 is a diagram for explaining detection of a range of vertex candidates.

[0011] Hereinafter, an example of an embodiment of the disclosed technology will be described with reference to the drawings.

[0012] Before describing the details of the embodiment, a volcano plot, which is the detection target of the detection device of the embodiment, will be described.

[0013] The upper diagram in Figure 1 shows a volcano plot of the decomposition of formic acid on a catalyst. This graph plots data for each material, with the explanatory variable, the heat of formation of metal formate (kJ / mol), on the X axis, and the target variable, the temperature (K) at which the reaction reaches a specific rate, on the Y axis. As shown in the lower diagram in Figure 1, this graph shows the adsorption reaction onto the catalyst, indicated by the solid line, and the desorption reaction from the catalyst, indicated by the dashed line. Because the reaction on the lower side of the graph (thick line portion) is dominant, the graph in the example in Figure 1 is mountain-shaped.

[0014] As shown in Figure 1, in the real world, there are cases where the positive and negative slopes of two related variables reverse at certain values. This occurs because the dominance relationship changes among variables that are composed of multiple causal factors. The peaks of the mountain-shaped graph shown in Figure 1 are important points where physical properties change. Note that volcano plots do not always have a mountain shape like that shown in Figure 1, and can also have a valley shape.

[0015] To obtain a volcano plot like the one above, it is necessary to specify which explanatory variables should be set on the X-axis. This requires specialized knowledge and rules of thumb, so in the past, experts would specify the variables by guessing from the data.

[0016] In addition, it may be difficult to grasp the trend of peaks or valleys from a scatter plot of plotted data. For example, a scatter plot may be wide in the Y-axis direction or have a large variance overall, making it difficult to detect a volcano plot.

[0017] Furthermore, it may be possible to detect volcano plots more easily by setting conditions for the plotted data in addition to the explanatory variables set on the X-axis. However, when there are a large number of explanatory variables, the number of possible conditions increases exponentially, and when such conditions are set manually, there is a risk that conditions may be overlooked, making comprehensive verification difficult.

[0018] Furthermore, if you increase the number of explanatory variables in the data, you will need to reconsider the specification of the explanatory variables on the X-axis. It may also become more difficult to detect a volcano plot from a scatter plot. This makes it difficult to handle an increase in explanatory variables.

[0019] Therefore, in this embodiment, characteristic conditions are extracted from the relationship between the conditions, which are combinations of explanatory variables, and the objective variable, and candidates for volcano plots are detected based on the extracted conditions. The detection system according to this embodiment will be described below.

[0020] 2, the detection system 100 according to this embodiment includes a detection device 10 and a user terminal 70. The detection device 10 and the user terminal 70 are connected via a network. Note that the number of user terminals 70 connected to one detection device 10 is not limited to one, and may be two or more.

[0021] The user terminal 70 is an information processing terminal used by a user who wishes to perform volcano plot detection on the target data. For example, if the target data is data on the physical properties of a material, the user terminal 70 is placed in a laboratory or the like. Also, if the target data is marketing data, the user terminal 70 is placed in a company's marketing department or the like.

[0022] The user terminal 70 has a data input function, a communication function, and a display function. The user terminal 70 transmits a data set of target data to the detection device 10. The user terminal 70 also receives screen data of the detection result transmitted from the detection device 10, and displays the detection result screen indicated by the screen data on the display device.

[0023] The detection device 10 is an information processing device such as a server device, and is located, for example, in a data center. In the above example, the detection device 10 may also be located in a laboratory, a company, or the like. The detection device 10 is not limited to performing detection processing on a data set transmitted from a user terminal 70 via a network. The detection device 10 may also perform detection processing on a data set directly input to the detection device 10, and display a detection result screen on a display device provided in the device itself.

[0024] 2, the detection device 10 functionally includes an extraction unit 12, a detection unit 14, and an output unit 16. A condition database (DB) 22 and a detection result DB 24 are stored in a predetermined storage area of ​​the detection device 10.

[0025] The extraction unit 12 acquires a dataset transmitted from the user terminal 70. The dataset includes multiple data, and each data includes one or more explanatory variables and a dependent variable. An example of the dataset is schematically shown in the upper diagram of FIG. 3. The extraction unit 12 extracts specific conditions that have a predetermined correlation with the dependent variable from among conditions that are combinations of explanatory variables included in the data. Specifically, the extraction unit 12 extracts, as specific conditions, conditions under which an index indicating the strength of the causal relationship between an explanatory variable and a dependent variable that are in a cause-and-effect relationship is equal to or greater than a predetermined reference value.

[0026] A more specific description will be given. As shown in the lower diagram of FIG. 3 , the extraction unit 12 extracts conditions that are correlated with the objective variable from among conditions that are exhaustive combinations of explanatory variables included in the dataset. The extraction unit 12 may determine whether each condition is correlated with the objective variable based on whether a statistical value, such as normalized mutual information, is equal to or greater than a predetermined reference value. In this case, for explanatory variables whose values ​​are numeric, the extraction unit 12 divides the range of values ​​into multiple values ​​and multi-values ​​the explanatory variable. For example, if the value of explanatory variable X1 is numeric, the multi-values ​​are set such that X1 > Xa and X1 ≦ Xa. Furthermore, for explanatory variables whose values ​​are categorical values, the extraction unit 12 multi-values ​​the explanatory variable for each type of categorical value. For example, if the category value of explanatory variable X9 is 0 or 1, the multi-values ​​are set such that X9 = 0 and X9 = 1. Note that the number of multi-values ​​is not limited to two and may be three or more.

[0027] Furthermore, as shown in FIG. 4 , the extraction unit 12 performs statistical causal estimation for each extracted condition and estimates an index indicating the strength of the causal relationship. For example, the extraction unit 12 generates a causal graph based on a subset obtained by extracting data corresponding to each condition from the dataset. The causal graph has each explanatory variable and each target variable as a node, and nodes having a causal relationship are connected by edges leading from the cause node to the result node. Furthermore, a causal effect, which is an index indicating the strength of the influence of a cause on a result, is assigned to each edge. The causal effect is an example of an index indicating the strength of the causal relationship. The extraction unit 12 statistically estimates the causal effect based on the subset corresponding to the condition.

[0028] When the causal effect (dotted line in FIG. 4 ) estimated for the explanatory variable that causes the dependent variable is equal to or greater than a predetermined reference value, the extraction unit 12 extracts the condition as a specific condition and stores the condition in the condition DB 22. In this way, by extracting specific conditions based on statistical causal estimation, a group (subset) that has a significant causal effect on the dependent variable under the specific condition is extracted. In other words, each specific condition can be said to include an explanatory variable that identifies each group.

[0029] The detection unit 14 extracts subsets of the data set that can be divided into a first subgroup and a second subgroup based on a specific condition, and detects subsets in which the positive and negative signs of the regression coefficients between the dependent variable and the explanatory variables change between the first subgroup and the second subgroup as candidates for a volcano plot.

[0030] Specifically, the detection unit 14 retrieves one specific condition from the condition DB 22. The detection unit 14 selects one explanatory variable from the explanatory variables included in the retrieved specific condition as a specific explanatory variable. The detection unit 14 extracts from the dataset a subset that satisfies the conditions of explanatory variables other than the specific explanatory variable included in the specific condition. The detection unit 14 plots data included in the subset in an area with the specific explanatory variable as the X-axis and the objective variable as the Y-axis. When the detection unit 14 divides the area into a first area and a second area at a position where the X-axis value is the threshold value of the specific explanatory variable, the detection unit 14 identifies the data included in the first area as a first subgroup and the data included in the second area as a second subgroup.

[0031] For example, suppose the detection unit 14 retrieves the specific condition, Condition 2: X1 > Xa ∧ X2 > Xb, from the condition DB 22 shown in the lower diagram of FIG. 4 and selects the explanatory variable X1 as the specific explanatory variable. In this case, the detection unit 14 extracts data satisfying the condition X2 > Xb from the dataset as a subset. Then, as shown in the upper diagram of FIG. 5, the detection unit 14 plots the data of the subset in an area with the specific explanatory variable X1 on the X axis and the objective variable on the Y axis. Furthermore, based on the specific explanatory variable condition X1 > Xa, the detection unit 14 sets the threshold value to X1 = Xa, and classifies the data in the area where X1 ≦ Xa as subgroup A and the data in the area where X1 > Xa as subgroup B. Note that subgroup A and subgroup B are examples of a first subgroup and a second subgroup.

[0032] As shown in the lower diagram of Fig. 5, the detection unit 14 obtains a regression line for each of subgroup A and subgroup B and calculates the regression coefficient (slope of the regression line). As shown in the lower diagram of Fig. 5, if the sign of the regression coefficient of subgroup A changes (is reversed) from the sign of the regression coefficient of subgroup B, the detection unit 14 detects the extracted subset as a candidate for a volcano plot. Furthermore, the detection unit 14 detects the intersection of the regression line of subgroup A and the regression line of subgroup B as a candidate for the apex of the detected candidate for a volcano plot.

[0033] When the detection unit 14 detects a volcano plot candidate, it stores the corresponding specific condition, the specific explanatory variable set on the X-axis, the regression equation showing the regression line of each subgroup, and the X-coordinate of the vertex in the detection result DB 24. FIG. 6 shows an example of the detection result DB 24. If the extracted specific condition includes other explanatory variables, the detection unit 14 performs the same detection process using the other explanatory variables as specific explanatory variables. Furthermore, the detection unit 14 performs the same detection process for all specific conditions stored in the condition DB 22.

[0034] The output unit 16 outputs the detection results stored in the detection result DB 24 to the user terminal 70. At this time, the output unit 16 may generate screen data showing a detection result screen based on the dataset and the detection results and output the screen data to the user terminal 70. The detection result screen may include, for example, as shown in the lower diagram of FIG. 5 , volcano plot candidates plotted in an area with the specific explanatory variable on the X axis, regression lines for the data of each of the first subgroup (subgroup A) and the second subgroup (subgroup B), and an indication of the vertices. When multiple detection results are stored in the detection result DB 24, the output unit 16 generates and outputs detection result screen data for each of the detection results.

[0035] The detection device 10 may be realized by, for example, a computer 40 shown in FIG. 7 . The computer 40 includes a CPU (Central Processing Unit) 41, a GPU (Graphics Processing Unit) 42, a memory 43 as a temporary storage area, and a non-volatile storage device 44. The computer 40 also includes an input / output device 45 such as an input device and a display device, and an R / W (Read / Write) device 46 that controls reading and writing of data from and to a storage medium 49. The computer 40 also includes a communication I / F (Interface) 47 that is connected to a network such as the Internet. The CPU 41, GPU 42, memory 43, storage device 44, input / output device 45, R / W device 46, and communication I / F 47 are connected to one another via a bus 48.

[0036] The storage device 44 is, for example, a hard disk drive (HDD), a solid state drive (SSD), or a flash memory. The storage device 44, which serves as a storage medium, stores a detection program 50 for causing the computer 40 to function as the detection device 10. The detection program 50 includes extraction process control instructions 52, detection process control instructions 54, and output process control instructions 56. The storage device 44 also has an information storage area 60 in which information constituting each of the condition DB 22 and the detection result DB 24 is stored.

[0037] The CPU 41 reads the detection program 50 from the storage device 44, loads it in the memory 43, and sequentially executes the control instructions of the detection program 50. The CPU 41 operates as the extraction unit 12 shown in FIG. 2 by executing the extraction process control instruction 52. The CPU 41 also operates as the detection unit 14 shown in FIG. 2 by executing the detection process control instruction 54. The CPU 41 also operates as the output unit 16 shown in FIG. 2 by executing the output process control instruction 56. The CPU 41 also reads information from the information storage area 60 and loads the condition DB 22 and the detection result DB 24 into the memory 43. As a result, the computer 40 that executed the detection program 50 functions as the detection device 10. The CPU 41 that executes the program is hardware. A portion of the program may also be executed by the GPU 42.

[0038] The functions realized by the detection program 50 may be realized by, for example, a semiconductor integrated circuit, more specifically, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or the like.

[0039] The hardware configuration of the user terminal 70 is generally similar to that of the detection device 10, and therefore a description thereof will be omitted.

[0040] Next, the operation of the detection system 100 according to this embodiment will be described. When a data set is transmitted from the user terminal 70 and input to the detection device 10, the detection device 10 executes the detection process shown in FIG. 8 . The user terminal 70 then receives screen data output from the detection device 10 by the detection process, and the user terminal displays a detection result screen indicated by the screen data on a display device. The detection process is an example of a detection method of the disclosed technology. The detection process will be described in detail below.

[0041] In step S10, the extraction unit 12 acquires the data set transmitted from the user terminal 70 and input to the detection device 10. Next, in step S12, the extraction unit 12 extracts conditions that are comprehensive combinations of explanatory variables included in the data set, and that have a correlation with the objective variable, such as a statistical value of normalized mutual information being equal to or greater than a predetermined reference value.

[0042] Next, in step S14, the extraction unit 12 estimates the causal effect of the explanatory variables that cause the target variable for each extracted condition using statistical causal estimation, extracts conditions whose causal effect is equal to or greater than a predetermined standard value as specific conditions, and stores them in the condition DB 22.

[0043] Next, in step S16, the detection unit 14 retrieves one specific condition from the condition DB 22, and selects one explanatory variable from the explanatory variables included in the retrieved specific condition as a specific explanatory variable. The detection unit 14 then extracts a subset from the data set that satisfies the conditions of explanatory variables other than the specific explanatory variable included in the specific condition, and plots the data included in the subset in an area with the specific explanatory variable on the X axis and the response variable on the Y axis.

[0044] Next, in step S18, when the detection unit 14 divides the region into a first region and a second region at the position where the X-axis value is the threshold value of the specific explanatory variable, the detection unit 14 identifies the data included in the first region as subgroup A (first subgroup) and the data included in the second region as subgroup B (second subgroup). Next, in step S20, the detection unit 14 calculates the regression coefficient of the specific explanatory variable with respect to the objective variable for each of subgroup A and subgroup B.

[0045] Next, in step S22, the detection unit 14 determines whether the signs of the regression coefficients of subgroup A and subgroup B are reversed. If they are reversed, the process proceeds to step S24, and if they are not reversed, the process proceeds to step S26. In step S24, the detection unit 14 stores the specific conditions, the specific explanatory variables set on the X-axis, the regression equations indicating the regression lines of each subgroup, and the X-coordinates of the vertices in the detection result DB 24.

[0046] If the specific condition includes other explanatory variables, the processes of steps S16 to S24 are repeated with the other explanatory variables as specific explanatory variables. Also, the processes of steps S16 to S24 are repeated for all specific conditions stored in the condition DB 22.

[0047] Next, in step S26, the output unit 16 generates screen data for the detection result screen based on the dataset and the detection results stored in the detection result DB 24, and outputs the generated screen data to the user terminal 70, and the detection process is terminated.

[0048] As described above, according to the detection system of this embodiment, the detection device extracts specific conditions that have a predetermined correlation with the dependent variable from among conditions that are combinations of explanatory variables in a plurality of data sets including one or more explanatory variables and a dependent variable. The detection device also extracts subsets of the plurality of data sets that are divided into a first subgroup and a second subgroup based on the specific conditions. The detection device then detects subsets in which the positive and negative signs of the regression coefficients between the dependent variable and the dependent variable change between the first subgroup and the second subgroup as candidates for a volcano plot. This makes it possible to detect a volcano plot under various conditions without relying on the knowledge and experience of an expert. For example, if the dataset is data related to the physical properties of a material, the detected volcano plot can be used to find points of change in the physical properties.

[0049] Furthermore, as specific conditions, conditions under which the causal effect estimated by statistical causal estimation is equal to or greater than a reference value are extracted. This allows conditions under which a subset having a special causal relationship with respect to the dependent variable can be identified as specific conditions. Furthermore, data plotted in an area with the specific explanatory variable included in the specific conditions as the X-axis is divided by a threshold value for the specific explanatory variable, and candidates for a volcano plot are detected based on whether the sign of the regression coefficient is reversed before and after the division. This is based on the idea that a volcano plot is likely to form when the sign of the regression coefficient is reversed before and after the threshold value for the explanatory variable that can identify a subset having a special causal relationship with respect to the dependent variable.

[0050] Furthermore, by extracting specific conditions from a comprehensive combination of explanatory variables, it is possible to detect volcano plots that take various conditions into consideration, reducing the chance of overlooking conditions compared to when conditions are set manually. As a result, it is easy to expand the search range, and for example, if the data set is data on the physical properties of materials, it can contribute to the discovery of new physical properties.

[0051] In addition, since the manual work can be significantly reduced, it becomes easier to add explanatory variables, etc.

[0052] An application example of this embodiment will now be described with reference to FIG.

[0053] In this application example, data including a target variable (Target Variable) and explanatory variables (Descriptor, Atom) was targeted. As shown in the upper diagram of Figure 9, in a scatter plot in which all data is plotted in an area with Descriptor as the X-axis without considering specific conditions, it is unclear whether a volcano plot is formed.

[0054] This embodiment was applied to this data set, and the following specific condition was extracted. Specific condition: Descriptor≧0.7 ∧ Atom=Ru Furthermore, under this specific condition, Descriptor was selected as the specific explanatory variable. In this case, the regression coefficients calculated for each subgroup separated before and after Descriptor=0.7 were as follows. Note that "'C'==1" indicates that condition C is met, and "'C'==0" indicates that condition C is not met. 'Descriptor≧0.7'==1 ∧ 'Atom=Ru'==1 → α=1.2622 'Descriptor≧0.7'==0 ∧ 'Atom=Ru'==1 → α=-1.4546

[0055] As mentioned above, the positive and negative signs of the regression coefficients are reversed before and after the threshold value of the specific explanatory variable, so this case was detected as a candidate for a volcano plot. The lower diagram in Figure 9 shows the detected volcano plot candidates. As shown in the lower diagram in Figure 9, the volcano plot was successfully detected when Atom = Ru.

[0056] In the above embodiment, the intersection of the regression lines of the subgroups is detected as a candidate for the apex of the volcano plot, but this is not limiting. The threshold value of the specific explanatory variable that separates the subgroups may also be detected as a candidate for the apex.

[0057] Furthermore, because the causal effect calculated when extracting specific conditions is a statistical estimate, the reliability of the estimation result can also be estimated. For example, the causal effect is calculated as the weight of each edge in the causal graph, and this weight is output with a range that takes error into account. That is, the higher the reliability of the estimation result, the narrower the range, and the lower the reliability, the wider the range. Therefore, a range obtained by multiplying this width by a predetermined coefficient and adding values ​​before and after the vertex may be detected as the range of vertex candidates. For example, if the threshold value of a specific explanatory variable is used as a vertex candidate, a range obtained by adding the above width x coefficient before and after the threshold may be detected as the range of vertex candidates, as shown in FIG. 10. Furthermore, if the intersection of regression lines is used as a vertex candidate, a range in the X-axis direction where the regions obtained by adding the above width x coefficient to each regression line overlap may be detected as the range of candidates, as shown in FIG. 11.

[0058] In the above embodiment, the detection program is stored (installed) in advance in a storage device, but this is not limiting. The program according to the disclosed technology may be provided in a form stored in a storage medium such as a CD-ROM, a DVD-ROM, or a USB memory.

[0059] REFERENCE SIGNS LIST 10 Detection device 12 Extraction unit 14 Detection unit 16 Output unit 22 Condition DB 24 Detection result DB 40 Computer 41 CPU 42 GPU 43 Memory 44 Storage device 45 Input / output device 46 R / W device 47 Communication I / F 48 Bus 49 Storage medium 50 Detection program 52 Extraction process control command 54 Detection process control command 56 Output process control command 60 Information storage area 70 User terminal 100 Detection system

Claims

1. A detection program for causing a computer to execute a process including: extracting, from a plurality of data sets including one or more explanatory variables and a dependent variable, specific conditions that have a predetermined correlation with the dependent variable from among conditions that are combinations of the explanatory variables; and detecting, from subsets of the plurality of data sets divided into a first subgroup and a second subgroup based on the specific conditions, subsets in which the positive and negative signs of the regression coefficients between the dependent variable and the explanatory variables change between the first subgroup and the second subgroup, as candidates for a volcano plot.

2. The detection program according to claim 1, wherein the process of extracting the specific conditions includes extracting conditions under which an index indicating the strength of the causal relationship between an explanatory variable in a cause-and-effect relationship and the objective variable is equal to or greater than a predetermined reference value.

3. The detection program according to claim 2, wherein the process of extracting the specific conditions includes extracting conditions that are combinations of explanatory variables that are correlated with the objective variable from among the comprehensive combinations of explanatory variables included in the plurality of data, and extracting conditions for the extracted conditions under which the index estimated by statistical causal estimation with the objective variable is equal to or greater than a predetermined reference value.

4. The detection program according to claim 2 or 3, wherein the process of detecting candidates for the volcano plot includes plotting a subset of the plurality of data that meets the conditions of explanatory variables other than a specific explanatory variable selected from explanatory variables included in the specific condition in an area with the specific explanatory variable as a first axis and the target variable as a second axis, and dividing the area into a first area and a second area at a position where the value of the first axis is the threshold value of the specific explanatory variable under the specific condition, so that the data included in the first area is the first subgroup and the data included in the second area is the second subgroup.

5. The detection program according to claim 4, wherein the process of detecting the candidate volcano plot includes detecting an intersection of a regression line for the data of the first subgroup and a regression line for the data of the second subgroup, or a threshold value of the specific explanatory variable, as a candidate vertex in the candidate volcano plot.

6. The detection program according to claim 5, wherein the process of detecting the vertex candidates includes detecting a range according to the reliability of the index as the range of the vertex candidates.

7. The detection program according to claim 5, for causing the computer to execute processing that further includes displaying a detection result screen including the volcano plot candidate plotted in the region, a regression line for each of the data of the first subgroup and the second subgroup, and the vertex candidate.

8. The detection program according to claim 7, wherein, when a plurality of conditions are extracted from the plurality of data, the program detects the volcano plot candidates for each of the plurality of conditions and for each case where each of the explanatory variables included in the conditions is set as the specific explanatory variable, and when a plurality of the volcano plot candidates are detected, the program displays the detection result screen for each of the volcano plot candidates.

9. A detection method in which a computer executes a process including: extracting, from a plurality of data sets including one or more explanatory variables and a dependent variable, specific conditions that have a predetermined correlation with the dependent variable from among conditions that are combinations of the explanatory variables; and detecting, from subsets of the plurality of data sets divided into a first subgroup and a second subgroup based on the specific conditions, subsets in which the positive and negative signs of the regression coefficients between the dependent variable and the explanatory variables change between the first subgroup and the second subgroup, as candidates for a volcano plot.

10. The detection method described in claim 9, wherein the process of extracting the specific conditions includes extracting conditions under which an index indicating the strength of the causal relationship between an explanatory variable in a cause-and-effect relationship and the objective variable is equal to or greater than a predetermined reference value.

11. The detection method described in claim 10, wherein the process of extracting the specific conditions includes extracting conditions that are combinations of explanatory variables that are correlated with the objective variable from among comprehensive combinations of explanatory variables included in the plurality of data, and extracting conditions under which the index estimated by statistical causal estimation with the objective variable is equal to or greater than a predetermined reference value for the extracted conditions.

12. The detection method according to claim 10 or 11, wherein the process of detecting candidates for the volcano plot includes plotting a subset of the plurality of data that meets the conditions of an explanatory variable other than a specific explanatory variable selected from explanatory variables included in the specific condition in an area with the specific explanatory variable as a first axis and the target variable as a second axis, and dividing the area into a first area and a second area at a position where the value of the first axis is the threshold value of the specific explanatory variable under the specific condition, so that the data included in the first area is the first subgroup and the data included in the second area is the second subgroup.

13. The detection method according to claim 12, wherein the process of detecting the candidate volcano plot includes detecting an intersection of a regression line for the data of the first subgroup and a regression line for the data of the second subgroup, or a threshold value of the specific explanatory variable, as a candidate apex of the candidate volcano plot.

14. The detection method according to claim 13, wherein the process of detecting the vertex candidates includes detecting a range according to the reliability of the index as the range of the vertex candidates.

15. The detection method according to claim 13, wherein the computer executes processing that further includes displaying a detection result screen including the volcano plot candidate plotted in the region, a regression line for each of the data of the first subgroup and the second subgroup, and the vertex candidate.

16. The detection method according to claim 15, wherein, when a plurality of conditions are extracted from the plurality of data, the volcano plot candidates are detected for each of the plurality of conditions and for each case where each of the explanatory variables included in the conditions is set as the specific explanatory variable, and when a plurality of the volcano plot candidates are detected, the detection result screen for each of the volcano plot candidates is displayed.

17. A detection device comprising: an extraction unit that extracts, from a plurality of data sets including one or more explanatory variables and a dependent variable, specific conditions that have a predetermined correlation with the dependent variable from among conditions that are combinations of the explanatory variables; and a detection unit that detects, from subsets of the plurality of data sets divided into a first subgroup and a second subgroup based on the specific conditions, subsets in which the positive and negative signs of the regression coefficients between the dependent variable and the explanatory variables change between the first subgroup and the second subgroup, as candidates for a volcano plot.

18. A detection device as described in claim 17, wherein the extraction unit extracts conditions under which an index indicating the strength of the causal relationship between an explanatory variable in a cause-and-effect relationship and the objective variable is equal to or greater than a predetermined reference value.

19. The detection device described in claim 18, wherein the extraction unit extracts conditions that are combinations of explanatory variables that are correlated with the objective variable from among the comprehensive combinations of explanatory variables included in the plurality of data, and extracts conditions under which the index estimated by statistical causal estimation with the objective variable is equal to or greater than a predetermined reference value from the extracted conditions.

20. The detection device described in claim 18 or 19, wherein the detection unit plots a subset of the plurality of data that meets the conditions of explanatory variables other than a specific explanatory variable selected from explanatory variables included in the specific condition in an area with the specific explanatory variable as a first axis and the target variable as a second axis, and when the area is divided into a first area and a second area at a position where the value of the first axis is the threshold value of the specific explanatory variable under the specific condition, the data included in the first area is the first subgroup and the data included in the second area is the second subgroup.

Images