Machine learning based repair forecasting

Patent metadata

Field	Value
Publication number	US-2019156298-A1
Application number	US-201715816778-A
Country	US
Kind code	A1
Filing date	Nov 17, 2017
Priority date	Nov 17, 2017
Publication date	May 23, 2019
Grant date	—

Abstract

A data processing system may include instructions stored in a memory and executed by a processor to categorize a plurality of systems into clusters using an unsupervised machine learning method to analyze repair data parameters of a historical dataset relating to the plurality of systems. The system may assign a repair forecast to each cluster, and may generate a system repair forecasting model using selected predictor variables, the historical data set, and the repair forecasts according to a supervised machine learning method. The selected predictor variables may correspond to a mathematical combination of operational data parameters in the historical dataset.

First claim

What is claimed is:

1 . A data processing system for generating repair forecasting models, comprising:
one or more processors; a memory including one or more digital storage devices; and a plurality of instructions stored in the memory and executable by the one or more processors to:
receive a historical dataset relating to each system of a plurality of systems, the historical dataset including a plurality of repair data parameters and a plurality of operational data parameters;
automatically categorize the plurality of systems into at least two clusters by using an unsupervised machine learning method to analyze the historical dataset with respect to the plurality of repair parameters;
assign a repair forecast label to each cluster of the at least two clusters; and
generate a system repair forecasting model, using one or more selected predictor variables, the repair forecast labels, and the historical dataset to train the system repair forecasting model according to a supervised machine learning method;
wherein each of the one or more selected predictor variables corresponds to a mathematical combination of one or more of the operational data parameters.

2 . The system of claim 1 , wherein the plurality of instructions are further executable by the one or more processors to:
receive an input corresponding to a selected subset of the plurality of repair parameters; and perform a first statistical analysis of differences between the at least two clusters with respect to the selected subset.

3 . The system of claim 2 , wherein the plurality of instructions are further executable by the one or more processors to:
report values resulting from the performed statistical analysis to a user; merge two or more clusters, according to a selection by the user; and repeat statistical analysis, reporting, and cluster merging until a final set of at least two clusters is confirmed by the user.

4 . The system of claim 1 , wherein each repair forecast label indicates a difference between the respective cluster and the other clusters with respect to one or more repair data parameters.

5 . The system of claim 1 , wherein the plurality of instructions are further executable by the one or more processors to:
display an inter-cluster comparative analysis of one or more of the operational data parameters; and obtain the one or more selected predictor variables from an outside source.

6 . The system of claim 1 , wherein the plurality of instructions are further executable by the one or more processors to:
receive respective values of each predictor variable relating to a first one of the systems; classify the first system into a predicted cluster of the at least two clusters by using the system repair forecasting model to analyze the values of each predictor variable; and communicate information relating to the predicted cluster to a repair forecasting program.

7 . The system of claim 1 , wherein each system of the plurality of systems comprises an aircraft, and the operational data parameters relate to flight data of the aircraft.

8 . The system of claim 1 , wherein the plurality of instructions are further executable by the one or more processors to:
generate a plurality of system repair forecasting models, using the one or more selected predictor variables and the historical dataset to train each system repair forecasting model according to a supervised leave-one-out cross validation machine learning method; and choose the repair demand forecast model having greatest accuracy on training test data.

9 . A computer implemented method of generating a repair forecasting model, comprising:
receiving a historical dataset relating to each system of a plurality of systems, the historical dataset including a plurality of repair data parameters and a plurality of operational data parameters; categorizing the plurality of systems into at least two clusters by using an unsupervised machine learning method to analyze the historical dataset with respect to the plurality of repair parameters; assigning a repair forecast label to each cluster of the at least two clusters; and generating a system repair forecasting model, using one or more selected predictor variables, the repair forecast labels, and the historical dataset to train the system repair forecasting model according to a supervised machine learning method; wherein each of the one or more selected predictor variables corresponds to a mathematical combination of one or more of the operational data parameters.

10 . The method of claim 9 , further comprising:
receiving an input corresponding to a selected subset of the plurality of repair parameters; and performing a first statistical analysis of differences between the at least two clusters with respect to the selected subset.

11 . The method of claim 10 , further comprising:
reporting values resulting from the performed statistical analysis to a user; merging two or more clusters, according to a selection by the user; and repeating statistical analysis, reporting, and cluster merging until a final set of at least two clusters is confirmed by the user.

12 . The method of claim 9 , wherein each repair forecast label indicates a difference between the respective cluster and the other clusters with respect to one or more repair data parameters.

13 . The method of claim 9 , further comprising:
displaying an inter-cluster comparative analysis of one or more of the operational data parameters; and obtaining the one or more selected predictor variables from an outside source.

14 . The method of claim 9 , further comprising:
receiving respective values of each predictor variable relating to a first one of the systems; classifying the first system into a predicted cluster of the at least two clusters by using the system repair forecasting model to analyze the values of each predictor variable; and communicating information relating to the predicted cluster to a repair forecasting program.

15 . The method of claim 9 , further comprising:
generating a plurality of system repair forecasting models, using the one or more selected predictor variables and the historical dataset to train each system repair forecasting model according to a supervised leave-one-out cross validation machine learning method; and choosing the repair demand forecast model having greatest accuracy on training test data.

16 . A computer program product for generating repair forecasting models, the computer program product comprising:
a non-transitory computer-readable storage medium having computer-readable program code embodied therewith, the computer readable program code configured to cause a data processing system to generate a repair forecasting model, the computer readable program code comprising: at least one instruction to receive a historical dataset relating to each system of a plurality of systems, the historical dataset including a plurality of repair data parameters and a plurality of operational data parameters; at least one instruction to automatically categorize the plurality of systems into at least two clusters by using an unsupervised machine learning method to analyze the historical dataset with respect to the plurality of repair parameters; at least one instruction to assign a repair forecast label to each cluster of the at least two clusters; and at least one instruction to generate a system repair forecasting model, using one or more selected predictor variables, the repair forecast labels, and the historical dataset to train the system repair forec

Assignees

• Boeing Co

Inventors

• Ethington James M
• Sturlaugson Liessman

Classifications

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N20/20

  Ensemble learning · CPC title

• G06N20/00

  Machine learning · CPC title