Module learnrdr :: Class RDRLearnerMDL

Class RDRLearnerMDL

rules_adhoc.RDRLearner --+
                         |
                        RDRLearnerMDL

Class for learning ripple-down rules from data using the minimum description length principle.

Instance Methods

[hide private]

learn(self, data_file, features_file, max_depth, c, pos_symbol, neg_symbol, min_length, max_length, regenerate)
Learn RDR from data using MDL and display the results.

source code

Ruleset

greedy_set_cover_mdl(self, data, pos_data, ruleset, possible_rules, datapoint_bits, covered_data)
Find greedy set cover using MDL.

source code

(list of Rules, Rule, integer)

find_possible_rules(self, features, data, depth, covered_data, pos_length, datapoint_bits)
Find possible rules given a set of data points and features.

source code

Ruleset

find_rules(self, data, features, depth)
Find rules from data.

source code

Inherited from rules_adhoc.RDRLearner: check_data, count_errors, data_for_rule, greedy_set_cover, pos_data_for_depth

Method Details

[hide private]

learn(self, data_file, features_file, max_depth, c, pos_symbol, neg_symbol, min_length, max_length, regenerate)

source code

Learn RDR from data using MDL and display the results.

Parameters:

data_file (string) - file containing the training data.
features_file (string) - file containing the features or "all" if all possible substrings of given lengths are to be generated.
max_depth (integer) - maximum depth of rules.
c (float) - the value of parameter c in MDL.
pos_symbol (string) - symbol signifying positive example in data.
neg_symbol (string) - symbol signifying negative example in data.
min_length (integer) - minimum length for substrings to be generated.
max_length (integer) - maximum length for substrings to be generated.
regenerate (boolean) - whether to regenerate possible rules after each time a rule is added to the ruleset

Overrides: rules_adhoc.RDRLearner.learn

greedy_set_cover_mdl(self, data, pos_data, ruleset, possible_rules, datapoint_bits, covered_data)

source code

Find greedy set cover using MDL.

Parameters:

data (dict) - data points and their classifications.
pos_data (dict) - data points to be classified by the rules and their classifications.
ruleset (L(Ruleset)) - rules already found.
possible_rules (list) - possible rules that can be used for classification.
datapoint_bits (float) - bits needed to encode one data point.
covered_data (dict) - data points already covered by a rule.

Returns: Ruleset

Ruleset that was found.

find_possible_rules(self, features, data, depth, covered_data, pos_length, datapoint_bits)

source code

Find possible rules given a set of data points and features.

Parameters:

data (dict) - data points and their classifications.
depth (integer) - maximum depth for exceptions.
covered_data (dict) - data points already covered by a rule.
pos_length (integer) - number of unclassified elements.
datapoint_bits (float) - bits needed to encode one data point.
features (list)

Returns: (list of Rules, Rule, integer)

Possible rules, best rule and its index.

Overrides: rules_adhoc.RDRLearner.find_possible_rules

find_rules(self, data, features, depth)

source code

Find rules from data.

Parameters:

data (dict) - data points and their classifications.
features (list) - possible features for classification
depth (integer) - maximum depth for exceptions for the rules to be found.

Returns: Ruleset

Ruleset that was found.

Overrides: rules_adhoc.RDRLearner.find_rules