Rule learning method, program, and device selecting rule for updating weights based on confidence value
A rule learning method in machine learning includes distributing features to a given number of buckets based on a weight of the features which are correlated with a training example; specifying a feature with a maximum gain value as a rule based on a weight of the training example from each of the buckets; calculating a confidence value of the specified rule based on the weight of the training example; storing the specified rule and the confidence value in a rule data storage unit; updating the weights of the training examples based on the specified rule, the confidence value of the specified rule, data of the training example, and the weight of the training example; and repeating the distributing, the specifying, the calculating, the storing, and the updating, when the rule and the confidence value are to be further generated.
1. A rule learning method, which makes a computer execute a rule learning processing in machine learning, the method comprising:
calculating weights of features, registered in a training example data storage unit storing a plurality of training examples correlated with one or a plurality of the features, based on a weight of each training example correlated with each of the features;
sorting the features in descending order of the weights of the features;
distributing the features to a given number of buckets in the descending order;
specifying a maximum gain value feature in a bucket as a specified rule, where a gain of each feature in the bucket is calculated based on the weight of each training example which includes that feature;
calculating a confidence value of the specified rule based on the weight of a correlated training example;
storing a combination of the specified rule and the confidence value in a rule data storage unit;
updating weights of the training examples based on the specified rule, the confidence value of the specified rule, data of the training examples, and the weights of the training examples; and
repeating the distributing, the specifying, the calculating, the storing, and the updating, when the rule and the confidence value are to be further generated after the updating is applied to all the buckets.
2. The rule learning method according to claim 1 , wherein the weight of each feature is a sum of the weights of the training examples correlated with the feature.
3. The rule learning method according to claim 1 , wherein, each training example is correlated with a label showing whether the training example is true or false, and the gain is calculated with respect to associated training examples correlated with a given feature by an absolute value of a difference between a square root of the sum of the weights of the associated training examples correlated with the label showing true and a square root of the sum of the weights of the training examples correlated with the label showing false.
4. A non-transitory storage medium storing a rule learning program, which when executed by a computer, causes the computer to perform a method, the method comprising:
calculating weights of features, registered in a training example data storage unit storing a plurality of the training examples correlated with one or a plurality of the features, based on a weight of each training example correlated with each of the features;
sorting the features in descending order of the weights of the features;
distributing the features to a given number of buckets in the descending order;
specifying a maximum gain value feature in a bucket as a specified rule, where a gain of each feature in the bucket is calculated based on the weight of each training example which includes that feature;
calculating a confidence value of the specified rule based on the weight of a correlated training example;
storing a combination of the specified rule and the confidence value in a rule data storage unit;
updating weights of the training examples based on the specified rule, the confidence value of the specified rule, data of the training examples, and the weights of the training examples; and
repeating the distributing, the specifying, the calculating, the storing, and the updating, when the rule and the confidence value are to be further generated after the updating is applied to all the buckets.
5. A rule learning device comprising:
a training example data storage unit which stores a plurality of training examples correlated with one or a plurality of the features, and the weight of each training example;
a processor which
calculates weights of features based on a weight of each training example correlated with each of the features which are registered in said training example data storage unit,
sorts the features in descending order of the weights of the features;
distributes the features to a given number of buckets in the descending order;
specifies a maximum gain value feature as a specified rule, where a gain of each feature in the bucket is calculated based on the weight of each training example which includes that feature, and
calculates a confidence value of the specified rule based on the weight of a correlated training example;
a rule data storage unit which stores a combination of the specified rule and the confidence value; and
an updating unit which updates weights of the training examples based on the specified rule, the confidence value of the specified rule, data of the training examples, and the weights of the training examples; and
a repeating unit which repeats the distributing, the specifying, the calculating, the storing, and the updating, when the rule and the confidence value are to be further generated after the updating is applied to all the buckets.