Process and system for developing a predictive model
First Claim
1. A computer implemented process for developing a model for business applications which predicts the value of a single dependent variable based on the value of at least one independent variable comprising the steps of:
- (a) providing a dataset containing a plurality of observations each containing a value for said dependent variable and values for said at least one independent variable;
(b) creating from said dataset of an initial generation of chromosomes each comprising a predictive model, said chromosome creating step comprising forming a chromosome having a plurality of observed variable segments each formed by a plurality of genes, a plurality of interaction segments each formed by a plurality of genes, and an intercept gene positioned after said plurality of interaction segments;
wherein each interaction segment consists of the following genes—
an include/exclude gene, a coefficient gene, a first variable gene, an operator gene and a second variable gene;
(c) determining a quantitative fitness measure for each chromosome in said initial generation; and
(d) creating a next generation of chromosomes by selecting a number of chromosomes from said initial generation, crossing said selected chromosomes by at least one of a cloning and a crossover technique, and mutating said chromosomes.
14 Assignments
0 Petitions
Accused Products
Abstract
The present invention relates to a computer implemented process for developing a model which predicts the value of a single dependent variable based on the value of at least one independent variable. The process comprises the steps of creating a dataset containing a plurality of observations each containing a value for the dependent variable and values for the at least one independent variable, creating from the dataset a plurality of original chromosomes each comprising a possible predictive model, developing a quantitative fitness measure for each chromosome, and creating a new generation of chromosomes by selecting a number of the original chromosomes based upon the fitness measures, crossing the selected original chromosomes by at least one of a cloning and a pure (standard) crossover technique, and mutating the crossed chromosomes. A system for carrying out the process of the present invention is also described.
44 Citations
43 Claims
-
1. A computer implemented process for developing a model for business applications which predicts the value of a single dependent variable based on the value of at least one independent variable comprising the steps of:
-
(a) providing a dataset containing a plurality of observations each containing a value for said dependent variable and values for said at least one independent variable; (b) creating from said dataset of an initial generation of chromosomes each comprising a predictive model, said chromosome creating step comprising forming a chromosome having a plurality of observed variable segments each formed by a plurality of genes, a plurality of interaction segments each formed by a plurality of genes, and an intercept gene positioned after said plurality of interaction segments; wherein each interaction segment consists of the following genes—
an include/exclude gene, a coefficient gene, a first variable gene, an operator gene and a second variable gene;(c) determining a quantitative fitness measure for each chromosome in said initial generation; and (d) creating a next generation of chromosomes by selecting a number of chromosomes from said initial generation, crossing said selected chromosomes by at least one of a cloning and a crossover technique, and mutating said chromosomes. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25)
-
-
26. A computer implemented process for developing a model for business applications which predicts the value of a single dependent variable based on the value of at least one independent variable comprising the steps of:
-
(a) providing a dataset containing a plurality of observations each containing a value for said dependent variable and values for said at least one independent variable; (b) creating from said dataset of an initial generation of chromosomes each comprising a predictive model, said chromosome creating step comprising forming a chromosome having a plurality of observed variable segments each formed by a plurality of genes, a plurality of interaction segments each formed by a plurality of genes, and an interceptor gene positioned after said plurality of interaction segments; wherein each interaction segment consists of the following genes—
an include/exclude gene, a coefficient gene, a first variable gene, an operator gene and a second variable gene;(c) determining a quantitative fitness measure for each chromosome in said initial generation; (d) creating a next generation of chromosomes by selecting a number of chromosomes from said initial generation, crossing said selected chromosomes by at least one of a cloning and a crossover technique, and mutating said chromosomes; and said fitness measure determining step comprising providing a scheduler module and a plurality of fitness evaluator modules and using said scheduler module to direct one of said chromosomes in said generation to be measure to an available one of said evaluator modules. - View Dependent Claims (27, 28, 29, 30, 31)
-
-
32. A system for creating a predictive model for business applications comprising:
-
means for creating an initial generation of chromosomes from a dataset with each chromosome in said initial generation comprising a predictive model, said chromosome creating step comprising forming a chromosome having a plurality of observed variable segments each formed by a plurality of genes, a plurality of interaction segments each formed by a plurality of genes, and an interceptor gene positioned after said plurality of interaction segments; wherein each interaction segment consists of the following genes—
an include/exclude gene, a coefficient gene, a first variable gene, an operator gene and a second variable gene;means for determining a quantitative fitness measure for each chromosome in said initial generation; means for selecting a number of said initial generation chromosomes, for crossing the selected initial generation chromosomes by at least one of a cloning and a pure (standard) crossover technique, and for mutating the crossed chromosomes into a next generation of chromosomes; and a computer and said initial generation chromosome creating means comprising at least one modeler module within said computer. - View Dependent Claims (33, 34, 35, 36, 37)
-
-
38. A chromosome for predicting a model for business applications comprising a plurality of observed variable segments, a plurality of interaction segments and an intercept gene and further comprising said intercept gene being positioned on said chromosome after said plurality of interaction segments;
wherein each interaction segment consists of the following genes—
an include/exclude gene, a coefficient gene, a first variable gene, an operator gene and a second variable gene.- View Dependent Claims (39, 40, 41, 42, 43)
Specification