# US 12,182,679 B2 · Causality estimation of time series via supervised learning > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,182,679 B2 - Document ID: US12182679B2 - Status: Granted - Kind: B2 - Application number: 17/044,783 - Filed: 2019-03-29 - Published: 2021-04-22 - Granted: 2024-12-31 - Earliest claimed: 2018-04-03 - Assignees: Nippon Telegraph and Telephone Corporation - Applicants: Nippon Telegraph and Telephone Corporation - Inventors: Yoichi Chikahara; Akinori Fujino - Web page: https://patentgrep.com/patent/US12182679B2 ## CPC - G06N 20/10: using kernel methods, e.g. support vector machines [SVM] - G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting - G06F 18/2431: Multiple classes - G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks - G06N 20/00: Machine learning - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract A technique for estimating a causal relation, which can solve problems of a conventional technique and which does not require preliminary setting of a regression model is disclosed. An embodiment of the present invention relates to a causal relation learning device including: a feature value calculation unit that receives a correct label of three or more classification labels related to a causal relation of time-series data and time-series data corresponding to the correct label and calculates a feature value of the time-series data; and a classifier learning unit that learns a classifier using a set of the feature value and the correct label so that an output of the classifier with respect to the feature value is a largest value of an output value of the correct label. ## Claims ### Claim 1 A computer-implemented method for estimating aspects of time-series data, the method comprising: generating two or more sets of time-series data for machine learning based on a normalized combination of linear time-series data and non-linear time-series data; receiving a correct label of three or more classification labels for machine learning, wherein the correct label represents a causal relation among the two or more sets of time-series data for machine learning, and the three or more classification labels comprise: a first set of time-series data is a cause of a second set of time-series data, the second set of time-series data is a cause of the first set of time-series data, or no causal relation between the first set of time-series data and the second set of time-series data; generating a set of feature values of respective sets of the two or more sets of time-series data; and training, based on a set of the generated feature values and the correct label, a classifier based on machine learning for classifying a causal relation of time series data, wherein a first output value of the classifier using the generated feature value is a maximum output value of the classifier using the correct label; and estimating, using the trained classifier, the causal relation among a plurality of sets of time-series data. ### Claim 2 (depends on 1) The computer-implemented method of claim 1, the method further comprising: when the received two or more sets of time-series data are two sets of time-series data, generating, by the trained classifier, a first set of estimated respective values of the three or more classification labels as output; and when the received two or more sets of time-series data includes three or more sets of time-series data, generating, by the trained classifier, an average value of a second set of estimated respective values of the three or more classification labels as output. ### Claim 3 (depends on 1) The computer-implemented method of claim 1, the method further comprising: receiving a plurality of sets of time-series data for estimating a causal relationship; and estimating, using the trained classifier, one of the three or more classification labels for the received plurality of sets of time-series data without preconfiguring a regression model for fitting the plurality of sets of time-series data. ### Claim 4 (depends on 1) The computer-implemented method of claim 1, the method further comprising: receiving a plurality of sets of time-series data; and determining, based on the plurality of sets of time-series data using the trained classifier, the correct classification label, wherein the correct classification label is either a set of output of the three or more classification labels or the output with the largest value. ### Claim 5 (depends on 1) The computer-implemented method of claim 1, wherein the first set of time-series data relates to trend data of foreign currency exchanges, and wherein the second set of the two or more sets of time-series data relates to trend data of stock prices. ### Claim 6 A system for estimating aspects of time-series data, the system comprises: a processor; and a memory storing computer-executable instructions that when executed by the processor cause the system to: generating two or more sets of time-series data for machine learning based on a normalized combination of linear time-series data and non-linear time-series data; receive a correct label of three or more classification labels for machine learning, wherein the correct label represents a causal relation among the two or more sets of time-series data for machine learning, and the three or more classification labels comprise: a first set of time-series data is a cause of a second set of time-series data, the second set of time-series data is a cause of the first set of time-series data, or no causal relation between the first set of time-series data and the second set of time-series data; receive the two or more sets of time-series data for machine learning; generate a set of feature values of respective sets of the two or more sets of time-series data; train, based on a set of the generated feature values and the correct label, a classifier based on machine learning for classifying a causal relation of time series data, wherein a first output value of the classifier using the generated feature value is a maximum output value of the classifier using the correct label; and estimate, using the trained classifier, the causal relation among a plurality of sets of time-series data. ### Claim 7 (depends on 6) The system of claim 6, the computer-executable instructions when executed further causing the system to: when the received two or more sets of time-series data are two sets of time-series data, generate, by the trained classifier, a first set of estimated respective values of the three or more classification labels as output; and when the received two or more sets of time-series data includes three or more sets of time-series data, generate, by the trained classifier, average values of a second set of estimated respective values of the three or more classification labels as output. ### Claim 8 (depends on 6) The system of claim 6, the computer-executable instructions when executed further causing the system to: receive a plurality of sets of time-series data for estimating a causal relationship; and estimate, using the trained classifier, one of the three or more classification labels for the received plurality of sets of time-series data without preconfiguring a regression model for fitting the plurality of sets of time-series data. ### Claim 9 (depends on 6) The system of claim 6, the computer-executable instructions when executed further causing the system to: receive a plurality of sets of time-series data; and determine, based on the plurality of sets of time-series data using the trained classifier, the correct classification label, wherein the correct classification label is either a set of output of the three or more classification labels or the output with the largest value. ### Claim 10 (depends on 6) The system of claim 6, wherein the first set of time-series data relates to trend data of foreign currency exchanges, and wherein the second set of the two or more sets of time-series data relates to trend data of stock prices. ### Claim 11 A computer-readable non-transitory recording medium storing computer-executable instructions that when executed by a processor cause a computer system to: generating two or more sets of time-series data for machine learning based on a normalized combination of linear time-series data and non-linear time-series data; receive a correct label of three or more classification labels for machine learning, wherein the correct label represents a causal relation among the two or more sets of time-series data for machine learning, and the three or more classification labels comprise: a first set of time-series data is a cause of a second set of time-series data, the second set of time-series data is a cause of the first set of time-series data, or no causal relation between the first set of time-series data and the second set of time-series data; receive the two or more sets of time-series data for machine learning; generate a set of feature values of respective sets of the two or more sets of time-series data; train, based on a set of the generated feature values and the correct label, a classifier based on machine learning for classifying a causal relation of time series data, wherein a first output value of the classifier using the generated feature value is a maximum output value of the classifier using the correct label; and estimate, using the trained classifier, the causal relation among a plurality of sets of time-series data. ### Claim 12 (depends on 11) The computer-readable non-transitory recording medium of claim 11, the computer-executable instructions when executed further causing the system to: when the received two or more sets of time-series data are two sets of time-series data, generate, by the trained classifier, a first set of estimated respective values of the three or more classification labels as output; and when the received two or more sets of time-series data includes three or more sets of time-series data, generate, by the trained classifier, average values of a second set of estimated respective values of the three or more classification labels as output. ### Claim 13 (depends on 11) The computer-readable non-transitory recording medium of claim 11, the computer-executable instructions when executed further causing the system to: receive a plurality of sets of time-series data for estimating a causal relationship; and estimate, using the trained classifier, one of the three or more classification labels for the received plurality of sets of time-series data without preconfiguring a regression model for fitting the plurality of sets of time-series data. ### Claim 14 (depends on 11) The computer-readable non-transitory recording medium of claim 11, the computer-executable instructions when executed further causing the system to: receive a plurality of sets of time-series data; and determine, based on the plurality of sets of time-series data using the trained classifier, the correct classification label, wherein the correct classification label is either a set of output of the three or more classification labels or the output with the largest value. ### Claim 15 (depends on 11) The computer-readable non-transitory recording medium of claim 11, wherein the first set of time-series data relates to trend data of foreign currency exchanges, and wherein the second set of the two or more sets of time-series data relates to trend data of stock prices.