JP6695431B2 - Analytical apparatus, analytical system and analytical method - Google Patents

Description

  The present invention relates to an analysis device, an analysis system and an analysis method for analyzing data.

  US Patent Application Publication No. 2004/011187 A1 is a computer-implemented method, system, and computer-readable storage for use with a clinical decision support system that identifies and provides information about a correlation between patient attributes and one or more adverse events (AEs). Disclose the medium. The process of U.S. Patent No. 6,096,981 processes database information including AEs and one or more patient attributes for correlation between AEs and patient attributes, one or more AEs and one or more patient attributes, Identifying at least one correlation between Correlations may be discovered via a correlation rule discovery process to determine one or more association rules. Each association rule meets certainty, support, and / or other thresholds. The process further provides the user with information or alerts based on the identified or discovered correlations.

  Patent Document 2 discloses a medical care support program that appropriately supports medical care. In the medical treatment support program of Patent Document 2, a treatment period of a patient for a diagnosed disease is compared with a reference cure period for the diagnosed disease, and when the treatment period of the patient exceeds the reference cure period, Searching for other diseases that develop similar symptoms to the diagnosed disease from the storage means that associates and stores the respective diseases that cause similar symptoms, and outputs the disease name information of the searched other diseases. , Causes the computer to execute the process.

Special table 2012-524945 gazette JP, 2014-199597, A

  However, the above-mentioned conventional technique has a problem that even if a learning model is generated from learning data, it is not known which factor is associated with which other factor. For example, when the objective variable is the disease probability and the factor is the dose of a plurality of drugs, for example, it is not known whether it is effective to administer a combination of drug A and drug B to a patient or whether side effects occur. There is.

  The present invention aims to analyze the effectiveness of a combination of factors.

  An analysis apparatus, an analysis system, and an analysis method according to one aspect of the invention disclosed in the present application, a storage device, a learning data set having a plurality of learning data including measured values of objective variables and measured values of a plurality of factors, A prediction data set having a plurality of prediction data derived from the learning data including prediction values of the plurality of factors, and a learning model indicating a relationship between the measured value of the objective variable and the measured values of the plurality of factors are stored. A first generation process of clustering the prediction data sets to generate a plurality of factor clusters so that the values of the plurality of factors are similar to each other, and using the prediction data set, A first calculation process for calculating a co-occurrence amount in which the plurality of factors co-occur by correlation, and clustering the plurality of factors based on the co-occurrence amount calculated in the first calculation process to obtain two or more factors. A second generation process for generating a plurality of co-occurrence clusters having one or more co-occurrence clusters including, and a specific factor cluster including two or more factors among the plurality of factor clusters generated by the first generation process. Of the predicted values of the two or more factors in the included specific prediction data group, the two or more specific factors indicated by the specific co-occurrence cluster among the plurality of co-occurrence clusters generated by the second generation processing are included. A second calculation process of calculating a predicted value of the objective variable in the specific factor cluster by giving the predicted value to the learning model.

  According to the exemplary embodiment of this invention, the effectiveness of the combination of factors can be analyzed. Problems, configurations and effects other than those described above will be clarified by the following description of the embodiments.

FIG. 1 is an explanatory diagram of a data analysis example according to the first embodiment. FIG. 2 is a block diagram showing a hardware configuration example of the analyzer. FIG. 3 is an explanatory diagram showing the detailed contents of the learning data shown in FIG. FIG. 4 is an explanatory diagram showing an example of the initial setting screen. FIG. 5 is a flowchart showing an example of an analysis processing procedure by the analysis device. FIG. 6 is an explanatory diagram showing the probability distribution of factors. FIG. 7 is an explanatory diagram showing an example of the integrated probability distribution. FIG. 8 is an explanatory diagram showing the result of factor clustering. FIG. 9 is an explanatory diagram showing a processing example of co-occurrence clustering. FIG. 10: is explanatory drawing which shows the prediction result by step S510. FIG. 11 is an explanatory diagram showing an example of a display screen. FIG. 12 is an explanatory diagram showing a system configuration example of the analysis system. FIG. 13 is a flowchart 1 showing an example of distributed processing procedure by the analysis system. FIG. 14 is a flowchart 2 showing an example of distributed processing procedure by the analysis system. FIG. 15 is a flowchart 3 showing an example of distributed processing procedure by the analysis system. FIG. 16 is a flowchart showing a modification of the flowchart 3 showing an example of the distributed processing procedure by the analysis system shown in FIG.

<Data analysis example>
FIG. 1 is an explanatory diagram of a data analysis example according to the first embodiment. (1) to (6) show the procedure of the analysis method by the analyzer. (1) The analysis device generates a learning model from the learning data set 10. In the learning data set 10, for example, the objective variable is the drug effect, specifically the disease probability, and the factor is the dose of a plurality of drugs to the patient. The disease probability can be expressed as 0% to 100%, but here, the disease is 1 (= 100%) and the health is 0 (= 0%). In addition, the factors are four explanatory variables of drug 1 to drug 4 for convenience, but actually, for example, tens of thousands to hundreds of millions of drugs. Each entry also indicates a patient. For convenience, the number of patients is 6 from A to F, but actually, for example, tens of thousands to hundreds of millions of patients.

  (1) In generating the learning model, the learning model generated includes a linear model and a non-linear model. Linear models include, for example, linear classification and logistic regression. Examples of the non-linear model include a neural network (Neural Network), a support vector machine (Support Vector Machine), an Adaboost, and a random forest (Random Forests). The user can select one of the models when generating the learning model. For example, the user may select a linear model when analyzing the effectiveness of a combination of factors at high speed, and may select a non-linear model when analyzing with high accuracy.

  (2) The analysis device generates the probability distribution 20 of each factor from the learning model generated in (1). Specifically, for example, the analyzer generates two sets of probability distributions 20 of factors derived from the learning data set 10 (referred to as d1 and d2, respectively) by using a probability sampling method represented by the Markov chain Monte Carlo method. .. Thereby, a large amount of virtual factor data can be collected.

  (3) The analyzer determines whether the probability distributions d1 and d2 of the factors generated in (2) converge to the same probability distribution. Specifically, for example, the Gelman-Rubin method is used for the convergence determination. Until it converges, the analyzer generates the probability distribution 20 of the factor (2).

  (4) The analysis device integrates the probability distributions d1 and d2 of the factors determined to converge in (3), and performs factor clustering on the integrated probability distribution of the factors (integrated probability distribution D). For the factor clustering, specifically, for example, k-means clustering is used. The number of clusters is set in advance. Here, the number of clusters is “3” as an example. As a result, in the factor clustering result 40, the entries of the integrated probability distribution D are classified into three types of patient types α, β, and γ.

  (5) Further, the analysis device executes co-occurrence clustering on the integrated probability distribution D. Specifically, for example, the analysis device calculates the correlation coefficient between the factors of the integrated probability distribution D as the co-occurrence amount. Then, the analysis device applies the hierarchical clustering method to the co-occurrence amount to generate a co-occurrence cluster. Here, it is assumed that the co-occurrence cluster 1 (drug 1, drug 2) and the co-occurrence cluster 2 (drug 3, drug 4) are obtained. Although the co-occurrence cluster is a combination of two factors here, it may be a combination of three or more factors.

  (6) The analyzer calculates the predicted value of the disease probability for each of the patient types α, β, γ by giving the learning model a factor belonging to the co-occurrence cluster for each of the patient types α, β, γ. In this way, the analyzer can analyze the effectiveness of the combination of factors.

<Example of hardware configuration of analyzer>
FIG. 2 is a block diagram showing a hardware configuration example of the analyzer. The analysis apparatus 200 has a processor 201, a storage device 202, an input device 203, an output device 204, and a communication interface (communication IF 205). The processor 201, storage device 202, input device 203, output device 204, and communication IF 205 are connected by a bus. The processor 201 controls the analysis device 200. The storage device 202 serves as a work area of the processor 201. The storage device 202 is a non-temporary or temporary recording medium that stores various programs and data. Examples of the storage device 202 include a ROM (Read Only Memory), a RAM (Random Access Memory), a HDD (Hard Disk Drive), and a flash memory. The input device 203 inputs data. Examples of the input device 203 include a keyboard, a mouse, a touch panel, a numeric keypad, and a scanner. The output device 204 outputs data. Examples of the output device 204 include a display and a printer. The communication IF 205 connects to a network and transmits / receives data.

<Example of learning data>
FIG. 3 is an explanatory diagram showing the detailed contents of the learning data set 10 shown in FIG. The learning data set 10 is, for example, data in a table format. In the following description of the database or table, the value of the AA field bbb (AA is a field name, bbb is a code) may be referred to as AAbbb. For example, the value of the patient ID field 301 is described as patient ID 301.

  The learning data set 10 has a patient ID field 301, an objective variable field 302, and a factor field 303. The values of the fields 301 to 303 in the same row form an entry that is patient information. In FIG. 3, the number of entries is “6”, but actually there are, for example, tens to hundreds of millions of patient entries.

  The patient ID field 301 is a storage area for storing a patient ID. The patient ID 301 is identification information that uniquely identifies the patient.

  The target variable field 302 is a storage area for storing a target variable for each patient ID 301. The objective variable 302 indicates a disease probability. The disease probability can be expressed as 0% to 100%, but since the learning data set 10 is an actual measurement value, the disease is 1 (= 100%) and the health is 0 (= 0%).

  The factor field 303 is a storage area that stores a plurality of factors. Factor 303 is an explanatory variable indicating the dose of the drug. In this example, the factors 303 are four explanatory variables, that is, drug 1 to drug 4 for convenience, but actually, for example, tens of thousands to hundreds of millions of drugs. The unit of the dose of the drug that is the factor 303 is determined for each drug.

  In FIG. 3, for the entry with the patient ID 301 of “patient A”, patient A is administered with drug 1 “20”, drug 2 “13.0”, and drug 4 “22.0”. Indicates a disease. In addition, for the entry having the patient ID 301 of “patient B”, the patient 1 is administered with medicine 1 “10”, medicine 2 “23.0”, medicine 3 “1”, and medicine 4 “31.0”. As a result, patient B is ill.

<Example of initial setting screen>
FIG. 4 is an explanatory diagram showing an example of the initial setting screen. The initial setting screen 400 is displayed on a display, which is an example of the output device 204, and is set by the input device 203. The machine learning selection area 401 is a pull-down interface for selecting a machine learning method. The factor clustering setting area 402 is an area for setting the clustering method and the number of clusters. The factor clustering selection area 403 is a pull-down interface for selecting a factor clustering method. The factor cluster number setting area 404 is an input field for setting the number of clusters desired to be obtained by factor clustering.

  The σ value setting area 405 is an input field for setting the σ value. The σ value is a fixed parameter used in the adoption rate α of the Markov chain Monte Carlo method in the generation of the probability distribution 20 of each factor (2) in FIG. The σ value is a value in the range of more than 0 and 1 or less.

  The co-occurrence clustering setting area 406 is an area for setting a co-occurrence method, a clustering method, the number of clusters, and a threshold value. The co-occurrence amount selection area 407 is a pull-down interface for selecting a method of calculating the co-occurrence amount. The co-occurrence clustering selection area 408 is a pull-down interface for selecting a method of co-occurrence clustering. The co-occurrence cluster number setting area 409 is an input field for setting the number of co-occurrence clusters to be obtained by factor clustering. The threshold value setting area 410 is an input field for setting a threshold value for a predicted value of a correlation value indicating the degree of association of factor clusters. The enter button 411 is a button for inputting the values of the items 401 to 410.

<Example of analysis processing procedure>
FIG. 5 is a flowchart showing an example of an analysis processing procedure by the analysis device 200. The analysis apparatus 200 executes the processing shown in the flowchart of FIG. 5 by causing the processor 201 to execute the analysis program stored in the storage device 202. First, the analysis device 200 executes initial setting (step S501). In the initial setting (step S501), the initial setting screen shown in FIG. 4 is displayed on the display. The user selects or inputs each item 401 to 409 on the initial setting screen. The analysis apparatus 200 reads the values of the items 401 to 409 by detecting the pressing of the input button 410.

  Next, the analysis device 200 generates a learning model from the learning data set 10 as shown in (1) of FIG. 1 (step S502). In the case of logistic regression, the learning model is expressed by the following equation (1).

y = f (x) = σ (w ^t x + b) (1)

y is a scalar indicating an objective variable. x is an m-dimensional feature vector. m corresponds to the number of factors. In the learning data set 10 of FIG. 3, the number of factors 303 is four (medicine 1 to drug 4), and therefore m = 4. σ () is a sigmoid function. The vector w and the scalar b are parameters of weight and bias, respectively, and are called learning parameters. In the case of a non-linear model, w ^t x in the sigmoid function σ () is replaced by a more complex function than w ^t x based on the vector w and the factor x.

  The analysis apparatus 200 selects a learning model according to the machine learning method selected in the machine learning selection area 401 of FIG. 4, and obtains a learning parameter expressing the learning model.

  Next, the analyzer 200 generates probability distributions d1 and d2 of the factors derived from the learning data set 10 by using the probability sampling method represented by the Markov chain Monte Carlo method, as shown in (2) of FIG. Yes (step S503).

  FIG. 6 is an explanatory diagram showing the probability distributions d1 and d2 of the factors. The factor probability distributions d1 and d2 have a virtual patient ID field 601, an objective variable field 602, and a factor field 603. The values of the fields 601 to 603 in the same row form an entry that is virtual patient information. The number of entries is the same as the number of entries in the learning data set 10.

  The virtual patient ID field 601 is a storage area for storing a virtual patient ID. The virtual patient ID 601 is identification information that uniquely identifies the virtual patient.

  The target variable field 602 is a storage area for storing a target variable for each virtual patient ID 601. The objective variable 602 indicates a disease probability. The disease probability is expressed as 0% to 100%.

  The factor field 603 is a storage area that stores a plurality of factors. The factor 603 is an explanatory variable indicating the dose of the drug. In this example, the number of factors 603 is the same as the number of factors 303 of the learning data set 10.

  An example of generation of virtual patient information that is entries of the factor probability distributions d1 and d2 will be described. The analysis apparatus 200 selects a factor vector of any entry from the entry group of the learning data set 10. For example, it is assumed that the factor vector x = (20,13.0,0,22.0) with the patient ID 301 of "patient A" is selected. The analysis apparatus 200 adds the random number value r to each element of the selected factor vector to obtain the virtual factor vector x ′ = (20 + r, 13.0 + r, 0 + r, 22.0 + r).

  The analysis apparatus 200 substitutes the selected factor vector x and the virtual factor vector x ′ into the equation (2) of the adoption rate α of the Markov chain Monte Carlo method.

  The function q is a Gaussian distribution function. The function q (x '| x) is a Gaussian distribution function indicating the probability of generating the virtual factor vector x'when the factor vector x is given. The function q (x | x ') is a Gaussian distribution function indicating the probability of generating the factor vector x when the virtual factor vector x'is given. The function f is, for example, the learning model generated in step S502 as shown in Expression (1). The σ value input to the σ value setting area 405 is substituted for σ. Depending on the σ value, the acceptance rate α has a Gaussian distribution including the patient information with the disease probability of (1−σ) or more. That is, the virtual factor vector x ′ of the virtual patient information having the disease probability of (1−σ) or more can be adopted at the adoption rate α.

  Next, when a uniform random number β is generated in the interval of 0 to 1 and the adoption rate α is equal to or greater than the threshold value β (for example, 1), the analysis device 200 adopts the virtual factor vector x ′. When the acceptance rate α is not greater than or equal to the threshold value, the analysis device 200 adopts the factor vector x. The adopted factor vector is expressed as adopted factor vector <x>.

  When the adoption rate α is equal to or larger than the threshold value β (for example, 1), the analysis device 200 compares the adoption factor vector <x> with the random number vector R. Specifically, for example, the analysis device 200 determines whether or not all the elements of the adoption factor vector <x> are greater than or equal to the corresponding elements of the random number vector R. When all the elements of the adoption factor vector <x> are equal to or more than the corresponding elements of the random number vector R, the analysis device 200 determines the adoption factor vector <x> as the virtual factor vector of the new virtual patient.

  When all the elements of the adopted factor vector <x> are not equal to or greater than the corresponding elements of the random number vector R, the analysis device 200 determines the factor vector x as the virtual factor vector of the new virtual patient. Although all the elements of the adoption factor vector <x> are greater than or equal to the corresponding elements of the random number vector R, the determination condition is that some elements of the adoption factor vector <x> are It may be more than the corresponding element.

  After that, the analysis apparatus 200 calculates the disease probability that is the objective variable 602 by giving the learning model a factor 603 that is a virtual factor vector of a new virtual patient in each virtual patient information entry. In this way, in step S503, the entry of the virtual patient information is set, and the probability distributions d1 and d2 of the factors are generated.

  Returning to FIG. 5, the analyzer 200 determines whether the probability distributions d1 and d2 of the factors converge to the same probability distribution as shown in (3) of FIG. 1 (step S504). Specifically, for example, the analysis device 200 calculates a convergence value for verifying whether the probability distributions d1 and d2 of the factors converge to the same probability distribution by the Gelman-Rubin method. More specifically, the analysis device 200 gives the column data of the probability distribution d1 of the factor and the column data of the probability distribution d2 of the factor corresponding to the column data to the Gelman-Rubin convergence determination formula to converge. Calculate the value Rhat.

  For example, the analysis apparatus 200 gives the column data of the objective variable 602 of the probability distribution d1 of the factor and the column data of the objective variable 602 of the probability distribution d2 of the factor to the Gelman-Rubin convergence determination formula to obtain the convergence value Rhat. calculate. Further, the analysis device 200 gives the column data of the drug 1 in the factor 603 of the factor probability distribution d1 and the column data of the drug 1 in the factor 603 of the factor probability distribution d2 to the Gelman-Rubin convergence determination formula, The convergence value Rhat is calculated. Similarly, the analyzer 200 calculates the convergence value Rhat for the column data after the medicine 2.

  If the convergence value Rhat is 1.1 or less, it is determined that the column data of the factor probability distributions d1 and d2 converge to the same probability distribution. The analysis device 200 deletes the column data determined not to converge. If the number of remaining column data is equal to or larger than the threshold value (for example, 50% or more), it means that the probability distributions d1 and d2 of the factors have converged to the same probability distribution (step S504: Yes), and the process proceeds to step S505. Transition. If the number of remaining column data is not equal to or more than the threshold value (step S504: No), the analysis device 200 returns to step S503 and regenerates the probability distributions d1 and d2 of the factors derived from the learning data set 10. Further, when even one column data of the factor 603 of the factor probability distributions d1 and d2 is deleted, the analyzer 200 gives the remaining factor 603 to the learning model and recalculates the objective variable 602.

  By deleting the column data that does not converge, it is possible to improve the reliability of the probability distributions d1 and d2 of the factors and improve the analysis accuracy. If the number of remaining column data is equal to or larger than the threshold value, the analysis device 200 may move to step S504 without deleting the column data determined not to converge. As a result, an analysis covering the factor 603 can be performed. Further, step S504 may not be executed. As a result, the analysis speed can be improved.

  Next, the analysis device 200 integrates the probability distributions d1 and d2 of the factors that are determined to converge in step S504 (step S505). The probability distribution of the integrated factors is referred to as an integrated probability distribution D.

  FIG. 7 is an explanatory diagram showing an example of the integrated probability distribution D. In FIG. 7, the content of the probability distributions d1 and d2 of the factors shown in FIG. 6 is connected for convenience of description. However, if any column data in the factor 603 is deleted in step S504, the integrated probability distribution Even in D, the state is deleted.

  Next, as shown in (4) of FIG. 1, the analysis apparatus 200 uses the integrated probability distribution D to generate a factor cluster by factor clustering (step S506). In the initial setting (step S501), the analyzer 200 executes the factor clustering selected in the factor clustering selection area 403 to generate the number of factor clusters set in the factor cluster number setting area 404.

  FIG. 8 is an explanatory diagram showing the result 40 of the factor clustering. The factor clustering result 40 has a patient type ID field 801, an objective variable field 802, and a factor field 803. The value of each field 801 to 803 in the same row constitutes an entry that becomes patient type information.

  The patient type ID field 801 is a storage area for storing a patient type ID. The patient type ID 801 is identification information that uniquely identifies the patient type classified by the factor clustering.

  The target variable field 802 is a storage area for storing a target variable for each patient type ID 801. The objective variable 802 indicates a disease probability. The disease probability is expressed as 0% to 100%.

  The factor field 803 is a storage area that stores a plurality of factors. Factor 803 is an explanatory variable indicating the dose of drug to the patient type. In the present example, the factor 803 is four explanatory variables of drug 1 to drug 4 for convenience, but actually, for example, the drug remains after the convergence determination (step S504).

  In FIG. 8, k-means clustering is used as the factor clustering, and the number of clusters is “3” as an example. As a result, the entries of the integrated probability distribution D are classified into the factor clusters of the three patient types α, β and γ.

  Returning to FIG. 5, the analyzer 200 calculates the statistical value of each factor from each factor cluster (step S507). Specifically, for example, the analysis device 200 sets a statistical value in the virtual patient information in the integrated probability distribution D belonging to the patient type of the entry in the factor field 803. The statistical value is, for example, a median value. In addition to the median value, the average value, the maximum value, the minimum value, or a randomly selected value may be used. Further, the analysis device 200 calculates the disease probability that is the objective variable 802 by giving the learning model a statistical value that is the factor 803. In this way, the patient type factor 803 and the explanatory variable 802 are aggregated into a statistical value and a statistical probability-derived disease probability.

  In addition, the analysis device 200 calculates the co-occurrence amount of the factors of the integrated probability distribution D (step S508). The co-occurrence amount is a correlation value between two factors. Specifically, for example, the analysis device 200 combines all factors in the integrated probability distribution D in a brute force manner and calculates a correlation value between the factors. The correlation value is calculated by the calculation method selected in the co-occurrence amount selection area 407 in the initial setting (step S501).

  Next, the analysis device 200 generates a co-occurrence cluster by co-occurrence clustering, as shown in (5) of FIG. 1 (step S509). Specifically, for example, the analysis device 200 applies a hierarchical clustering method to the co-occurrence amount to generate a co-occurrence cluster. Hierarchical clustering is to set individual data as one co-occurrence cluster, calculate the similarity between co-occurrence clusters, merge the most similar co-occurrence clusters, and all co-occurrence clusters are one cluster. It is a clustering that repeats the process until and becomes a dendrogram. Here, the similarity between co-occurrence clusters is, for example, the short distance between co-occurrence clusters. Specifically, the distance between co-occurrence clusters is defined by, for example, the nearest neighbor method, the farthest neighbor method, or the center of gravity method.

  FIG. 9 is an explanatory diagram showing a processing example of co-occurrence clustering (S508, S509). (A) shows the process of step S508. The co-occurrence amount table 900 is a table that holds correlation values between factors. (B) shows the process of step S509. In (B), the analyzer 200 deletes the correlation value of the same factor. Further, the analysis device 200 converts the correlation value into a correlation value obtained by subtracting the correlation value from 1 for hierarchical clustering. In (B), the smaller the correlation value, the more similar the factors are. Therefore, the analysis device 200 selects the combination of factors having the smallest correlation value as the co-occurrence cluster. In the case of (B), a combination of drug 1 and drug 2 (co-occurrence cluster 1) and a combination of drug 3 and drug 4 (co-occurrence cluster 2) are selected. Although the co-occurrence cluster is a combination of two factors here, it may be a combination of three or more factors.

  The process (B) is executed until the number of co-occurrence clusters reaches the number of co-occurrence clusters set in the co-occurrence cluster number setting area 409, or until no more clusters can be merged. ..

  Returning to FIG. 5, the analysis device 200 calculates the predicted value of the co-occurrence cluster as shown in (6) of FIG. 1 (step S510). Specifically, for example, the analysis device 200 gives the learning model the factors belonging to the co-occurrence cluster for each of the patient types α, β, γ, to thereby predict the disease probability for each of the patient types α, β, γ. To calculate.

  FIG. 10: is explanatory drawing which shows the prediction result 1000 by step S510. In this way, the analysis device 200 can analyze the effectiveness of the combination of factors.

  Returning to FIG. 5, the analysis device 200 executes threshold processing of the prediction result 1000 (step S511). Specifically, for example, the analysis device 200 selects a combination of a patient type and a factor cluster whose predicted value is equal to or higher than a threshold value. For example, when the threshold value set in the threshold value setting area 410 is “0.8”, the analysis device 200 determines that the factor cluster 1 of the patient type α, the factor cluster 1 of the patient type β, and the patient type γ. Select factor cluster 1 as the calculation marker.

  The analyzer 200 outputs the processing result of step S510 or S511 (step S512). Specifically, for example, the analysis apparatus 200 controls the display screen of the display which is an example of the output device 204 to display the processing result on the display screen, or transmits the processing result to the external device via the communication IF 205. Alternatively, the processing result is written in the storage device 202. Also, the convergence determination result of step S504 may be output.

<Display screen example>
FIG. 11 is an explanatory diagram showing an example of a display screen. The display screen 1100 is displayed on a display that is an example of the output device 204. The display screen 1100 has a score display area 1101, a prediction result display area 1102, and a dendrogram display area 1103. In the score display area 1101, the convergence value Rhat in the convergence determination (step S504) is displayed. In the prediction result display area 1102, the prediction result 1000 shown in FIG. 10 is displayed. As shown in FIG. 11, a bar graph may be displayed. A dendrogram in hierarchical clustering is displayed in the dendrogram display area 1103. In this way, the intermediate results and final results of the processing shown in FIG. 5 are displayed on the display screen 1100.

  As described above, according to the first embodiment, the analysis device 200 clusters the prediction data set (for example, the integrated probability distribution D) so that the values of the plurality of factors are similar to each other to generate a plurality of factor clusters. The first generation process is executed (step S506). The analysis apparatus 200 uses the prediction data set (for example, the integrated probability distribution D) to execute the first calculation process of calculating the co-occurrence amount in which a plurality of factors co-occur due to the correlation of a plurality of factors (step S508). .. The analysis device 200 clusters a plurality of factors based on the co-occurrence amount calculated by the first calculation process to generate a plurality of co-occurrence clusters having one or more co-occurrence clusters including two or more factors. Generation processing is executed (step S509). The analysis device 200 selects the first predicted value of the two or more factors in the specific predicted data group included in the specific factor cluster including the two or more factors among the plurality of factor clusters generated by the first generation process. The predicted values of two or more specific factors indicated by a specific co-occurrence cluster among a plurality of co-occurrence clusters generated by the 2 generation process are given to the learning model. Then, the analysis device 200 executes the second calculation process of calculating the predicted value of the objective variable in the specific factor cluster (step S510).

  Thereby, the analysis device 200 can analyze the effectiveness of the combination of factors based on the predicted value of the objective variable in the specific factor cluster in which a plurality of factors co-occur.

  Further, the analysis device 200 executes a third calculation process of calculating a statistical value representing the predicted values of the two or more factors in the specific factor cluster, based on the predicted values of the two or more factors in the specific predicted data group. (Step S510). As a result, the analyzer 200 can reduce the amount of calculation when calculating the predicted value of the target variable in the specific factor cluster in which a plurality of factors co-occur. Therefore, the analysis speed can be improved.

  The analysis apparatus 200 also executes a setting process for setting the type of learning model (step S501). In addition, the analysis device 200 uses the measured values of the objective variables and the measured values of the plurality of factors to generate a learning model of the type set by the setting process, and executes a third generation process of storing the learning model in the storage device. Yes (step S502). This allows the user to select the type of learning model according to the purpose.

  In the setting process, the analysis device 200 sets a linear model or a non-linear model as the type. As a result, the analysis apparatus 200 can improve the analysis speed when the linear model is set, and can improve the analysis accuracy when the nonlinear model is set. In other words, the user can select a linear model if he / she wants to obtain an analysis result earlier and a non-linear model if he / she wants to improve the analysis accuracy.

  Further, the prediction data set (for example, integrated probability distribution D) may be a data set generated from the learning data set 10 by the probability sampling method using the learning model. As a result, the predicted data set (for example, the integrated probability distribution D) becomes a data set that depends on the learning model. Therefore, for example, when a non-linear model is set, the prediction data set (for example, integrated probability distribution D) becomes a data set with higher accuracy than when a linear model is set.

  In addition, the analysis device 200 adopts either one of the prediction data or the data similar to the prediction data by the probability sampling method (for example, Markov chain Monte Carlo method) using the learning model, so that two prediction data groups (for example, , The probability generation distributions d1, d2) of the factors are executed (step S503). The data similar to the prediction data is, as described above, data in which a random value is added to each value of the factor that is the prediction data. The analysis apparatus 200 executes the determination process of determining whether or not the two prediction data groups (for example, the probability distributions d1 and d2 of factors) generated by the fourth generation process converge to the same probability distribution (step). S504). The analysis apparatus 200 integrates two prediction data groups (for example, the probability distributions d1 and d2 of factors) based on the determination result of the determination process to generate a prediction data set (for example, the integrated probability distribution D). The process is executed (step S505).

  By the determination process, it is determined whether or not the two prediction data groups (for example, the probability distributions d1 and d2 of the factors) converge to the same probability distribution, for example, the probability distribution of the learning data set 10. As a result, if the two prediction data groups (for example, the probability distributions d1 and d2 of factors) are similar to the learning data set 10 if they converge, the two prediction data groups (for example, the probability distribution d1 of the factor). , D2), a predicted data set (for example, integrated probability distribution D) is generated. This makes it possible to improve the certainty of the predicted data set (for example, the integrated probability distribution D) as a predicted value, that is, the generation accuracy.

  In addition, the analysis device 200 adopts a probability sampling method (for example, Markov chain Monte Carlo method) using a learning model to adopt either prediction data or data similar to the prediction data, and the value of a parameter that controls the adoption rate α ( For example, a setting process for setting the σ value) is executed (step S501). As a result, it is possible to adopt a factor that is an objective variable of (1-σ) or more at the adoption rate α.

  In addition, the analysis device 200 executes a setting process for setting the number of generated factor clusters (step S501). Thereby, the analysis device 200 can generate the number of factor clusters designated by the user. Specifically, for example, as the number of generated factor clusters increases, the prediction data set (for example, integrated probability distribution D) is subdivided. Thus, the user can set the number of generated factor clusters to a lower number when the analysis result is desired to be obtained earlier, and can set the number of generated factor clusters to a higher number when the analysis accuracy is desired to be improved.

  The analysis apparatus 200 also executes a setting process for setting the number of co-occurrence clusters to be generated (step S501). As a result, the analysis apparatus 200 can generate the number of co-occurrence clusters designated by the user. Specifically, for example, as the number of generated co-occurrence clusters increases, the number of co-occurring factors and the number of combinations of co-occurring factors increase. Therefore, the user can set the number of generated co-occurrence clusters to a lower number if the analysis result is to be obtained earlier, and can set the number of generated co-occurrence clusters to a higher number to improve the analysis accuracy.

  Further, in Example 1, a plurality of factors 303 and 603 are set as doses of a plurality of drugs to a patient, and objective variables 302 and 602 are values indicating drug efficacy when a plurality of doses are administered to a patient (for example, Disease probability). This makes it possible to predict how much each of a plurality of drugs is administered to a patient of which type (factor cluster) and how much the drug is effective.

  In addition, in Example 1 described above, the drug efficacy analysis was described as an example, but it is also applicable to product recommendation. In this case, in the learning data set 10 shown in FIG. 3, the patient ID 301 is, for example, a customer instead of a patient. The factor 303 indicates, for example, the number of purchases (in the case of a product) or the number of uses (in the case of a service) of a product or service (the genre of the product or service may be used). The objective variable 302 indicates, for example, the purchase price (in the case of a product) or the usage price (in the case of a service) of a product or service (the genre of the product or service may be used). The same applies to the probability distributions d1 and d2 of the factors and the integrated probability distribution D.

  Further, in the case of analysis of news articles, in the learning data set 10 shown in FIG. 3, the patient ID 301 is replaced with, for example, news articles published in newspapers, magazines, and web pages instead of patients. The factor 303 indicates the number of appearances of a word, for example. The objective variable 302 indicates the genre of news articles such as politics, society, sports, and weather. The same applies to the probability distributions d1 and d2 of the factors and the integrated probability distribution D.

  Example 2 will be described. In the first embodiment, the analysis process shown in FIG. 5 is executed by one computer, but in the second embodiment, the analysis process shown in FIG. 5 is distributed by a plurality of computers. This will reduce the load on the computer and increase the analysis speed. Specifically, each computer has, for example, the hardware configuration shown in FIG.

  FIG. 12 is an explanatory diagram showing a system configuration example of the analysis system. The analysis system 1200 includes a plurality of computers (hereinafter, simply nodes) N0 to Nn (n is an integer of 1 or more) and one or more client terminals C. A plurality of nodes N0 to Nn (n is an integer of 2 or more) and one or more client terminals C are communicably connected via a network 1201. The node N0 is the master node N0, and the nodes N1 to Nn are the worker nodes N1 to Nn. The master node N0 manages the worker nodes N1 to Nn. The worker nodes N1 to Nn execute processing according to the instruction from the master node N0. Note that any one of the worker nodes N1 to Nn may be responsible for the function of the master node N0.

<Example of distributed processing procedure>
13 to 15 are flowcharts showing an example of distributed processing procedure by the analysis system 1200. Here, as an example, n = 2, that is, the analysis system 1200 is the master node N0, the worker nodes N1 and N2, and the client terminal C.

  First, the client terminal C executes initial setting (step S501) (step S1301). Then, the client terminal C transmits an analysis request, which is the setting content of the initial setting (step S501), to the master node N0 (step S1302).

  The master node N0 transmits a learning model generation request to the worker node N1 (step S1303). When receiving the learning model generation request, the worker node N1 generates a learning model as in step S502 (step S1304). After generating the learning model, the worker node N1 transmits the learning model to the master node N0 (step S1305). Upon receiving the learning model from the worker node N1, the master node N0 transmits the learning model to another worker node N2 (step S1306).

  Next, the master node N0 transmits a generation request of the factor probability distribution d1 to the worker node N1 (step S1307), and transmits a generation request of the factor probability distribution d2 to the worker node N2 (step S1308). Thereby, the probability distributions d1 and d2 of the factors can be generated by parallel processing.

  Next, the worker node N1 generates the probability distribution d1 of the factors derived from the learning data set 10 by using the probability sampling method represented by the Markov chain Monte Carlo method as in step S503 (step S1309). Similarly to step S503, the worker node N2 also uses the probability sampling method represented by the Markov chain Monte Carlo method to generate the probability distribution d2 of the factors derived from the learning data set 10 (step S1310). The worker node N1 transmits the generated probability distribution d1 of the factors to the master node N0 (step S1311). The worker node N2 also transmits the generated probability distribution d2 of the factors to the master node N0 (step S1312).

  Similar to step S504, the master node N0 determines whether the probability distributions d1 and d2 of the factors converge to the same probability distribution (step S1313). The master node N0 transmits the determination result to the client terminal C (step S1314). As shown in FIG. 11, the client terminal C receives and displays the determination result (for example, Gelman-Rubin score) (step S1315).

  In FIG. 14, the master node N0 integrates the probability distributions d1 and d2 of the factors to generate the integrated probability distribution D, as in step S505 (step S1401). Then, the master node N0 transmits a factor clustering request to the worker node N1 (step S1402). When the worker node N1 receives the factor clustering request, similarly to step S506, the worker node N1 uses the integrated probability distribution D to generate a factor cluster by factor clustering (step S1403). Further, the worker node N1 calculates the statistical value of each factor from each factor cluster, similarly to step S507 (step S1404). The worker node N1 transmits the calculated statistical value to the master node N0 (step S1405). The master node N0 transmits the received statistical value to another worker node N2 (step S1406).

  The master node N0 transmits a co-occurrence amount calculation request to the worker node N2 (step S1407). The worker node N2 calculates the co-occurrence amount of the factors of the integrated probability distribution D, as in step S508 (step S1408). Then, the worker node N2 transmits the calculated co-occurrence amount (see (A) of FIG. 9) to the master node N0 (step S1409).

  In FIG. 15, the master node N0 generates co-occurrence clusters by co-occurrence clustering, as in step S509, and generates co-occurrence cluster ID lists A and B (step S1501). The co-occurrence cluster ID list A is an ID list that uniquely identifies one of the entry groups obtained by dividing the entries of the integrated probability distribution D. The co-occurrence cluster ID list B is an ID list that uniquely identifies the other entry group obtained by dividing the entries of the integrated probability distribution D.

  The master node N0 transmits the ID list A of the co-occurrence cluster to the worker node N1 (step S1502) and the ID list B of the co-occurrence cluster to the worker node N2 (step S1503). The worker node N1 generates a co-occurrence cluster for the ID list A by co-occurrence clustering, similarly to step S509 (step S1504). Similarly to step S509, the worker node N2 also generates a co-occurrence cluster for the ID list B by co-occurrence clustering (step S1505).

  The worker node N1 calculates the predicted value of the co-occurrence cluster obtained in step S1504, as in step S510 (step S1506). Similarly to step S510, the worker node N2 also calculates the predicted value of the co-occurrence cluster obtained in step S1505 (step S1507). The worker node N1 saves the predicted value obtained in step S1506 in the storage device 202 (step S1508). The worker node N2 also stores the predicted value obtained in step S1507 in the storage device 202 (step S1509). The worker node N1 transmits the predicted value obtained in step S1506 to the master node N0 (step S1510). The worker node N2 also transmits the predicted value obtained in step S1507 to the master node N0 (step S1511).

  The master node N0 executes threshold value processing of the predicted value, similarly to step S511 (step S1512). Then, the master node N0 transmits the calculation marker which is the execution result to the client terminal C (step S1513). The client terminal C displays the calculation marker on the display screen (step S1514).

  FIG. 16 is a flowchart showing a modification of the flowchart 3 showing an example of the distributed processing procedure by the analysis system 1200 shown in FIG. In FIG. 15, the worker nodes N1 and N2 execute co-occurrence clustering in parallel for each of the ID lists A and B, thereby realizing high-speed processing. On the other hand, in FIG. 16, the co-occurrence cluster calculation of the ID lists A and B is executed by the master node N0, not by the worker nodes N1 and N2. The same steps as those in FIG. 15 are designated by the same step numbers, and the description thereof will be omitted.

  In FIG. 16, the master node N0 generates a co-occurrence cluster by co-occurrence clustering for the ID list A, as in step S509 (step S1602). The master node N0 transmits the co-occurrence cluster of the ID list A to the worker node N1 (step S1603).

  The worker node N1 calculates the predicted value of the co-occurrence cluster obtained in step S1602, as in step S510 (step S1604). The worker node N1 stores the predicted value obtained in step S1604 in the storage device 202 (step S1604). The worker node N1 transmits the predicted value obtained in step S1604 to the master node N0 (step S1606).

  Similar to step S509, the master node N0 generates co-occurrence clusters for the ID list B by co-occurrence clustering (step S1607). The master node N0 transmits the co-occurrence cluster of the ID list B to the worker node N2 (step S1608).

  The worker node N2 calculates the predicted value of the co-occurrence cluster obtained in step S1607, as in step S510 (step S1609). The worker node N1 stores the predicted value obtained in step S1609 in the storage device 202 (step S1610). The worker node N2 transmits the predicted value obtained in step S1609 to the master node N0 (step S1611).

  Thus, according to the second embodiment, the same effect as that of the first embodiment can be obtained. Further, according to the second embodiment, the analysis processing shown in FIG. 5 is distributed and processed by a plurality of computers. This makes it possible to reduce the load on the computer and increase the analysis speed. The distributed processing shown in FIGS. 13 to 16 is an example. Therefore, in addition to this, for example, at least two or more of the steps shown in FIGS. 13 to 16 may be executed by different computers.

  The present invention is not limited to the above-described embodiments, but includes various modifications and equivalent configurations within the spirit of the appended claims. For example, the above-described embodiments have been described in detail in order to explain the present invention in an easy-to-understand manner, and the present invention is not necessarily limited to those having all the configurations described. Further, part of the configuration of one embodiment may be replaced with the configuration of another embodiment. Further, the configuration of another embodiment may be added to the configuration of one embodiment. Moreover, you may add, delete, or replace another structure with respect to a part of structure of each Example.

  Further, each of the above-mentioned configurations, functions, processing units, processing means, etc. may be realized by hardware, for example, by designing a part or all of them with an integrated circuit, and the processor realizes each function. It may be realized by software by interpreting and executing the program.

  Information such as programs, tables, and files that implement each function can be stored in a storage device such as a memory, a hard disk, an SSD (Solid State Drive), or a recording medium such as an IC card, an SD card, or a DVD.

  Further, the control lines and information lines are shown to be necessary for explanation, and not all the control lines and information lines necessary for mounting are shown. In reality, it can be considered that almost all configurations are connected to each other.

Claims (12)

An analysis apparatus comprising: a processor that executes a program; and a storage device that stores the program,
The storage device includes a learning data set including a plurality of learning data sets each including an actual measurement value of an objective variable and an actual measurement value of a plurality of factors, and a prediction including a plurality of prediction data-derived prediction data including prediction values of the plurality of factors. A data set, and a learning model showing a relationship between the measured values of the objective variable and the measured values of the plurality of factors are stored,
The processor is
A first generation process of clustering the prediction data set so that the values of the plurality of factors are similar to each other, and generating a plurality of factor clusters;
A first calculation process of calculating a co-occurrence amount in which the plurality of factors are co-occurring by the correlation of the plurality of factors using the prediction data set;
A second generation process for clustering the plurality of factors based on the co-occurrence amount calculated by the first calculation process to generate a plurality of co-occurrence clusters having one or more co-occurrence clusters including two or more factors; ,
Of the predicted values of the two or more factors in the specific prediction data group included in the specific factor cluster including the two or more factors among the plurality of factor clusters generated by the first generation processing, the second generation Prediction of the target variable in the specific factor cluster by giving the learning model prediction values of two or more specific factors indicated by a specific co-occurrence cluster among the plurality of co-occurrence clusters generated by the processing. A second calculation process for calculating a value,
An analyzer for performing the following. The analysis device according to claim 1, wherein
The processor is
Executing a third calculation process for calculating a statistical value representative of the predicted values of the two or more factors in the specific factor cluster, based on the predicted values of the two or more factors in the specific predicted data group,
In the second calculation process, the processor includes two or more specific factors indicated by the specific co-occurrence cluster among the statistical values representing the predicted values of the two or more factors calculated by the third calculation process. The analysis device is characterized in that a predicted value of the objective variable in the specific factor cluster is calculated by giving the statistical value of the above to the learning model. The analysis device according to claim 1, wherein
The processor is
Setting processing for setting the type of the learning model,
A third generation process of generating a learning model of the type set by the setting process using the measured values of the objective variable and the measured values of the plurality of factors and storing the learning model in the storage device;
An analyzer for performing the following. The analysis device according to claim 3, wherein
In the setting process, the processor sets a linear model or a non-linear model as the type. The analysis device according to claim 1, wherein
The predictive data set is a data set generated from the learning data set by a probability sampling method using the learning model. The analysis device according to claim 1, wherein
The processor is
A fourth generation process for generating two prediction data groups by adopting one of the prediction data or data similar to the prediction data by a probability sampling method using the learning model;
Determination processing for determining whether or not the two prediction data groups generated by the fourth generation processing converge to the same probability distribution,
Performing an integration process of generating the prediction data set by integrating the two prediction data groups based on the determination result of the determination process,
In the first generation processing, the processor clusters the prediction data set obtained by the integration processing so that the values of the plurality of factors are similar to each other, and generates the plurality of factor clusters,
In the first calculation process, the processor calculates a co-occurrence amount in which the plurality of factors co-occur by correlation of the plurality of factors, using the prediction data set obtained by the integration process. Analyzer. The analysis device according to claim 6, wherein
The processor is
By performing a setting process for setting the value of a parameter that controls the adoption rate to adopt one of the prediction data or the data similar to the prediction data by the probability sampling method using the learning model,
In the fourth generation processing, the processor generates the two prediction data groups by adopting one of the prediction data or data similar to the prediction data based on the adoption rate. Analyzer. The analysis device according to claim 1, wherein
The processor is
Execute the setting process to set the number of generation of the factor cluster,
In the first generation processing, the processor clusters the prediction data set so that the values of the plurality of factors are similar to each other, and generates the number of generation factor clusters set by the setting processing. Analyzer. The analysis device according to claim 1, wherein
The processor is
Execute a setting process to set the number of co-occurrence clusters generated,
In the second generation process, the processor clusters the plurality of factors based on the co-occurrence amount calculated in the first calculation process, and has one or more co-occurrence clusters including two or more factors. An analyzing apparatus, wherein the number of generated clusters set by the setting process is generated. The analysis device according to claim 1, wherein
The analyzer is characterized in that the plurality of factors are doses of a plurality of drugs to a patient, and the objective variable is a value indicating a drug effect when the plurality of drugs are administered to the patient in the doses. An analysis system in which a plurality of computers are communicably connected,
Any one of the plurality of computers, a learning data set having a plurality of learning data including the actual measurement value of the objective variable and the actual measurement value of the plurality of factors, and the prediction data derived from the learning data including the prediction values of the plurality of factors Storing a prediction data set having a plurality of, and a learning model showing the relationship between the actual measurement value of the objective variable and the actual measurement value of the plurality of factors,
One of the plurality of computers,
A first generation process of clustering the prediction data set so that the values of the plurality of factors are similar to each other, and generating a plurality of factor clusters;
A first calculation process of calculating a co-occurrence amount in which the plurality of factors are co-occurring by the correlation of the plurality of factors using the prediction data set;
A second generation process for clustering the plurality of factors based on the co-occurrence amount calculated by the first calculation process to generate a plurality of co-occurrence clusters having one or more co-occurrence clusters including two or more factors; ,
Of the predicted values of the two or more factors in the specific prediction data group included in the specific factor cluster including the two or more factors among the plurality of factor clusters generated by the first generation processing, the second generation Prediction of the target variable in the specific factor cluster by giving the learning model prediction values of two or more specific factors indicated by a specific co-occurrence cluster among the plurality of co-occurrence clusters generated by the processing. A second calculation process for calculating a value,
An analysis system characterized by executing. An analysis method by an analysis device having a processor that executes a program and a storage device that stores the program,
The storage device includes a learning data set having a plurality of learning data including measured values of objective variables and measured values of a plurality of factors, and prediction having a plurality of prediction data derived from the learning data including predicted values of the plurality of factors. A data set, and a learning model showing a relationship between the measured values of the objective variable and the measured values of the plurality of factors are stored,
The processor is
A first generation process of clustering the prediction data set so that the values of the plurality of factors are similar to each other, and generating a plurality of factor clusters;
A first calculation process of calculating a co-occurrence amount in which the plurality of factors are co-occurring by the correlation of the plurality of factors using the prediction data set;
A second generation process for clustering the plurality of factors based on the co-occurrence amount calculated by the first calculation process to generate a plurality of co-occurrence clusters having one or more co-occurrence clusters including two or more factors; ,
Of the predicted values of the two or more factors in the specific prediction data group included in the specific factor cluster including the two or more factors among the plurality of factor clusters generated by the first generation processing, the second generation Prediction of the target variable in the specific factor cluster by giving the learning model prediction values of two or more specific factors indicated by a specific co-occurrence cluster among the plurality of co-occurrence clusters generated by the processing. A second calculation process for calculating a value,
An analysis method characterized by executing.

Images