IPC분류정보
국가/구분 |
United States(US) Patent
등록
|
국제특허분류(IPC7판) |
|
출원번호 |
US-0356505
(2009-01-20)
|
등록번호 |
US-8185486
(2012-05-22)
|
발명자
/ 주소 |
|
출원인 / 주소 |
|
대리인 / 주소 |
|
인용정보 |
피인용 횟수 :
15 인용 특허 :
107 |
초록
▼
A computer program product, method and system for transforming data into predictive models. The transformation of data into predictive models comprises a multi stage learning process that uses a plurality of algorithms at each stage to select output for use in the next stage. The final predictive mo
A computer program product, method and system for transforming data into predictive models. The transformation of data into predictive models comprises a multi stage learning process that uses a plurality of algorithms at each stage to select output for use in the next stage. The final predictive model is a linear or nonlinear predictive model. Analyses of the model and the variables associated with it can be used to produce graphs and other management reports.
대표청구항
▼
1. A predictive model method, comprising: using a computer to perform at least one of the steps of:receiving first input data into a plurality of different types of initial predictive models to develop an initial model configuration by selecting an input data set from the plurality of predictive mod
1. A predictive model method, comprising: using a computer to perform at least one of the steps of:receiving first input data into a plurality of different types of initial predictive models to develop an initial model configuration by selecting an input data set from the plurality of predictive models using a variable selection algorithm after a training of each predictive model type is completed;receiving the input data set from said initial model configuration as an inputs into a second, model stage to develop an improvement to said initial model configuration and input data set as an output; andreceiving said second model stage output as an input into a third predictive model stage to develop and output a final predictive model where all the input data represents a physical object or substance and where the final predictive model consists of a linear predictive model or a nonlinear predictive model. 2. The method of claim 1, wherein said second model stage comprises an algorithm selected from the group consisting of LaGrange, Bayesian and path analysis. 3. The method of claim 1, wherein the data input to the second stage of processing further comprises the first input data, data not included in said first input data, and a combination thereof. 4. The method of claim 1, wherein the input data set from said initial model configuration comprises the input data to said initial model configuration after the training and a model selection has been completed. 5. The method of claim 1, further comprising: using a plurality of independent subpopulations to evolve a plurality of candidate predictive models with a plurality of genetic algorithms to identify a set of one or more changes that will optimize the final predictive model output value for one or more criteria. 6. The method of claim 1, wherein the types of initial predictive models are selected from the group consisting of CART; projection pursuit regression; generalized additive model (GAM), redundant regression network; boosted Naïve Bayes Regression; MARS; linear regression; and stepwise regression. 7. The method of claim 1, wherein the input data set from the plurality of predictive models is analyzed as required to identify one or more data clusters, the second stage analyzes the data for each cluster as required to develop and output a summary and the third predictive model stage analyzes said summaries as required to develop and output a final predictive model for each data cluster. 8. A computerized apparatus, comprising: means for receiving first input data into a plurality of different types of initial computerized predictive models to develop an initial model configuration by selecting an input data set from the plurality of predictive models using a variable selection algorithm after a training of each predictive model type is completed;means for receiving the input data set from said initial model configuration and a second input data as inputs into a second, model stage to develop an improvement to said initial model configuration as an output, said second input data comprising one of said first input data, data not included in said first input data, and a combination thereof; andmeans for receiving said second model stage output as an input into a third predictive model stage to develop and output a final predictive model where the final predictive model consists of a linear predictive model or a nonlinear predictive model. 9. The apparatus of claim 8, wherein said second model stage comprises an induction algorithm that receives the second input data and the input data set from the initial model configuration and transforms said inputs into a summary comprising the second model stage output where the induction algorithm is selected from the group consisting of entropy minimization, LaGrange, Bayesian and path analysis. 10. The apparatus of claim 8, wherein the input data set from said initial model configuration comprises the input data to said initial model configuration after the training and a model selection has been completed and where all the input data represents a physical object or substance. 11. The apparatus of claim 8, further comprising: means for using a plurality of independent subpopulations to evolve a plurality of candidate predictive models with a plurality of genetic algorithms to identify a set of one or more changes that will optimize the final predictive model output value for one or more criteria. 12. The apparatus of claim 8, wherein the types of initial computerized predictive models are selected from the group consisting of CART; projection pursuit regression; generalized additive model (GAM), redundant regression network; boosted Naïve Bayes Regression; MARS; linear regression; and stepwise regression. 13. The apparatus of claim 8, wherein the apparatus further comprises: means for analyzing the input data set from the plurality of predictive models as required to identify one or more data clusters, means for analyzing the data for each cluster as required to develop and output a summary and means for analyzing said summaries as required to develop and output a final predictive model for each data cluster. 14. A machine-readable medium tangibly embodying a program of non-transitory, machine-readable instructions executable by a digital processing apparatus to complete data transformation steps, comprising: receiving first input data into a plurality of different types of initial predictive models to develop an initial model configuration by selecting an input data set from the plurality of predictive models using a variable selection algorithm after a training of each predictive model type is completed;receiving the input data set from said initial model configuration and a second input data as inputs into a second, model stage to develop an improvement to said initial model configuration as an output, said second input data comprising one of said first input data, data not included in said first input data, and a combination thereof; andreceiving said second model stage output as an input into a third predictive model stage to develop and output a final predictive model where the final predictive model consists of a linear predictive model or a nonlinear predictive model. 15. The machine readable medium of claim 14, wherein said second model stage comprises an induction algorithm that receives the second input data and the input data set from the initial model configuration and transforms said inputs into a summary comprising the second model stage output where the induction algorithm is selected from the group consisting of entropy minimization, LaGrange, Bayesian and path analysis. 16. The machine readable medium of claim 14, wherein all input data represents a physical object or substance and the input data set from said initial model configuration comprises the input data to said initial model configuration after the training and a model selection has been completed. 17. The machine readable medium of claim 14, wherein the data transformation steps further comprise: using a plurality of independent subpopulations to evolve a plurality of candidate predictive models with a plurality of genetic algorithms to identify a set of one or more changes that will optimize the final predictive model output value for one or more criteria. 18. The machine readable medium of claim 14, wherein the types of initial predictive models are selected from the group consisting of CART; projection pursuit regression; generalized additive model (GAM), redundant regression network; boosted Naïve Bayes Regression; MARS; linear regression; and stepwise regression. 19. The machine readable medium of claim 14, wherein the input data set from the plurality of predictive models is analyzed as required to identify one or more data clusters, the second stage analyzes the data for each cluster as required to develop and output a summary and the third predictive model stage analyzes said summaries as required to develop and output a final predictive model for each data cluster and where the data clusters are identified by using a clustering algorithm selected from the group consisting of: unsupervised “Kohonen” neural network, neural network, decision tree, support vector method, K-nearest neighbor, expectation maximization (EM) and the segmental K-means algorithm. 20. The machine readable medium of claim 14, wherein the digital processing apparatus comprises a computer and all the input data represents a physical object or substance. 21. A non-transitory computer program product tangibly embodied on a computer readable medium and comprising a program code for directing at least one computer to: receive first input data into a plurality of different types of initial predictive models to develop an initial model configuration by selecting an input data set from the plurality of predictive models using a variable selection algorithm after a training of each predictive model type is completed;receive the input data set from said initial model configuration as an input into a second, model stage to develop an improvement to said initial model configuration and input data set as an output; andreceive said second model stage output as an input into a third predictive model stage to develop and output a final predictive model where the final predictive model consists of a linear predictive model or a nonlinear predictive model. 22. The computer program product of claim 21, wherein the final predictive model comprises a single, quantitative output variable and a plurality of input variables where said final predictive model identifies a contribution of each of the input variables to the output variable. 23. The computer program product of claim 22, wherein program code further directs the at least one computer to: quantify a dependence between each pair of input variables and output at least one report derived from one or more analyses of the predictive model and the input and output variables associated with said predictive model where the at least one report comprises a graphical report or a textual report for at least one scenario. 24. The computer program product of claim 21, wherein the types of initial computerized predictive models are selected from the group consisting of classification and regression tree; projection pursuit regression; generalized additive model (GAM), redundant regression network; multivariate adaptive regression splines; linear regression; and stepwise regression where the final predictive model type is selected from the same group. 25. The computer program product of claim 21, wherein the input data set from the initial model configuration is analyzed to identify one or more data clusters and a final predictive models is optionally developed for each of the data clusters.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.