Multistage machine learning process
First Claim
1. A method of operating an expert system to diagnose service affecting conditions on lines in a telecommunications network comprising:
- providing a set of training data, the data set made up of records with each record containing parameters indicative of measurements made on a line and a state of the line at the time the measurements were made;
selecting a plurality of subsets of the training data, wherein at least a first one of the subsets represents training data gathered during a first interval and at least a second one of the subsets representing training data gathered during a second interval after the first interval;
building a plurality of classification models, one from each of said plurality of subsets of the training data;
building a plurality of dependability models, one for each of said plurality of classification models;
providing test data derived from a line in the telecommunication network to each of said dependability models;
selecting a dependability model indicating the highest dependability from said plurality of dependability models; and
classifying the test data with the classification model associated with the dependability model indicating the highest dependability providing additional training data during a third interval after the second interval;
using the additional training data to create additional classification models and additional dependability models;
replacing the classification models and dependability models created with a first of the subset of data with the additional classification models and additional dependability models; and
thereafter applying test data on a line in the telecommunication network gathered after the third interval to the dependability models to select a classification model.
2 Assignments
0 Petitions
Accused Products
Abstract
The present invention includes a mechanism for applying expert knowledge and machine-learning routines to a continuous stream of information. The present method comprises learning a set of dependability models, one for each classification model, that characterize the situations in which each of the classification models is able to make correct predictions. At appropriate intervals the method produces new fault localization knowledge in the form of decision-tree based classification and dependability models. Such knowledge is used to enhance the existing classification knowledge already available. Each of these classification models has a particular sub-domain where it is the most reliable, and hence the best choice to use. For future unlabeled examples, these dependability models are consulted to select the most appropriate classification model, and the prediction of that classification model is then accepted.
81 Citations
21 Claims
-
1. A method of operating an expert system to diagnose service affecting conditions on lines in a telecommunications network comprising:
-
providing a set of training data, the data set made up of records with each record containing parameters indicative of measurements made on a line and a state of the line at the time the measurements were made;
selecting a plurality of subsets of the training data, wherein at least a first one of the subsets represents training data gathered during a first interval and at least a second one of the subsets representing training data gathered during a second interval after the first interval;
building a plurality of classification models, one from each of said plurality of subsets of the training data;
building a plurality of dependability models, one for each of said plurality of classification models;
providing test data derived from a line in the telecommunication network to each of said dependability models;
selecting a dependability model indicating the highest dependability from said plurality of dependability models; and
classifying the test data with the classification model associated with the dependability model indicating the highest dependability providing additional training data during a third interval after the second interval;
using the additional training data to create additional classification models and additional dependability models;
replacing the classification models and dependability models created with a first of the subset of data with the additional classification models and additional dependability models; and
thereafter applying test data on a line in the telecommunication network gathered after the third interval to the dependability models to select a classification model. - View Dependent Claims (2, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 21)
-
-
3. A method of operating an expert system comprising:
-
providing a set of training data;
selecting a plurality of subsets of the training data;
building a plurality of classification models, one from each of said plurality of subsets of the training data;
building a plurality of dependability models, one for each of said plurality of classification models;
providing test data to each of said dependability models;
selecting a dependability model indicating the highest dependability from said plurality of dependability models; and
classifying the test data with the classification model associated with the dependability model indicating the highest dependability;
wherein said step of building a plurality of dependability models comprises building decision tree based dependability models.
-
-
19. A method of operating an expert system comprising:
-
providing a set of training data;
selecting a plurality of subsets of the training data;
building a plurality of classification models, one from each of said plurality of subsets of the training data;
building a plurality of dependability models, one for each of said plurality of classification models;
providing test data to each of said dependability models;
selecting a dependability model indicating the highest dependability from said plurality of dependability models; and
classifying the test data with the classification model associated with the dependability model indicating the highest dependability;
wherein said step of building a plurality of dependability models includes the step of computing a dependability training set from said initial training set. - View Dependent Claims (20)
-
Specification