Pattern recognition using generalized association rules

A system for predicting an unknown value of an attribute of interest of a given item from a population of items, each item in the population having a plurality of variable attributes including the attribute of interest, comprising:

• an input device, which receives data indicative of values of at least some of the attributes of the given item;
• a memory, which stores association rules regarding the population, the association rules including at least one generalized association rule, each association rule comprising one or more conditions on one or more respective attribute values of the items predictive of the value of the attribute of interest, and the at least one generalized rule comprising a logical combination of such conditions using at least one logical operation from a group consisting of disjunction and negation in defining the conditions of the rule;
• and
• a processor, which receives the data from the input device and reads the association rules from the memory, and which applies the association rules including the at least one generalized association rule to the values included in the data so as to predict the unknown value of the attribute of interest and which generates an output responsive to the prediction.

A system for determining association rules for prediction of an unknown value of an attribute of interest of a given item from a population of items, each item in the population having a plurality of variable attributes including the attribute of interest, comprising:

• an input device, which receives data indicative of values of attributes of a training sample of items within the population including the attribute of interest;
• a memory, which stores the values of the attributes;
• and
• a computer, which reads the values from the memory and determines association rules regarding the population, the association rules including at least one generalized association rule, each association rule comprising one or more conditions on one or more respective attribute values of the items predictive of the value of the attribute of interest, and the at least one generalized rule comprising a logical combination of such conditions using at least one logical operation from a group consisting of disjunction and negation in defining the conditions of the rule, and which stores the association rules in the memory.

A system for determining association rules for prediction of an unknown value of an attribute of interest of a given item from a population of items, each item in the population having a plurality of variable attributes including the attribute of interest, comprising:

• an input device, which receives data indicative of values of attributes of a training sample of items within the population including the attribute of interest;
• a memory, which stores the values of the attributes;
• and
• a computer, which reads the values from the memory and determines simple association rules regarding the population, each simple association rule comprising one or more conditions on one or more respective attribute values of the items predictive of the value of the attribute of interest, such that if the simple association rule includes more than one such condition, the conditions are combined using a logical conjunction operation in defining the conditions of the rule, and which stores the association rules in the memory,
• wherein substantially all simple association rules applicable to the sample having respective probability and support greater than predetermined minimum values thereof are determined.

A system for determining association rules for prediction of an unknown value of an attribute of interest of a given item from a population of items, each item in the population having a plurality of variable attributes including the attribute of interest, comprising:

• an input device, which receives data indicative of values of attributes of a training sample of items within the population including the attribute of interest;
• a memory, which stores the values of the attributes;
• and
• a computer, which reads the values from the memory and determines association rules regarding the population, each association rule comprising one or more conditions on one or more respective attribute values of the items predictive of the value of the attribute of interest, and which stores the association rules in the memory,
• wherein the computer determines the association rules by constructing one or more contingency tables, each of whose entries corresponds to the number of items in the sample having a given value of the attribute of interest and satisfying a given, respective condition on one or more of the attributes other than the attribute of interest.