Machine learning

A method of classifying a case comprising:

• applying the case to each classifier of a first set of classifiers operable to generate predictions for categories, wherein the case includes one document, payroll data, web surfing data, and user profiles;
• applying the case to each classifier of a second set of classifiers also operable to generate predictions for the categories;
• applying outputs of a plurality of classifiers in the first set of classifiers to one or more classifiers in the second set of classifiers wherein the ithoutput of the first set classifiers does not input to the ithinput of the second set classifiers;
• determining a category in which the case belongs based on the output of at least one of the second set of classifiers so as to classify the case under at least one of the categories to organize the case.

An apparatus comprising:

• a plurality of classifiers in a first set, each classifier being operable to predict a category;
• a plurality of classifiers in a second set, each classifier in the second set corresponding to a classifier in the first set such that a classifier in the second set predicts the same category as a corresponding classifier in the first set;
• a training set input into the first set of classifiers, wherein the training set includes attributes of cases and labels;
• and
• each classifier in the second set receiving predictions from at least some of non-corresponding classifiers in the first set such that each classifier in the second set is trained using the predictions from the at least some of the non-corresponding classifiers and the training set wherein the ithoutput of the first set classifiers does not input to the ithinput of the second set classifiers;
• wherein the apparatus operates to classify documents, payroll data, web surfing data, user profiles, or any combination thereof, under different categories based on predictions provided by the first and second set of classifiers for classifying the documents, payroll data, web surfing data, user profiles, or any combination thereof.

A computer program provided on a computer readable medium and comprising code that when executed causes a computer to perform the following;

• training a first set of classifiers to predict a plurality of categories using attributes of training cases, wherein the training cases include documents, payroll data, web surfing data, user profiles, or any combination thereof;
• and
• training a second set of classifiers to predict the plurality of categories using the attributes of the training cases and using the predictions from one or more of the classifiers in the first set as inputs for training the second set of classifiers wherein the ithoutput of the first set classifiers does not input to the ithinput of the second set classifiers;
• receiving new cases for classification, wherein the new cases include one of the new documents, new payroll data, new web surfing data, and new user profiles;
• and
• classifying the new cases into the plurality of categories with the trained first set and second set of classifiers so as to classify each of the new cases under at least one topic as represented by one of the plurality of categories.