Prediction of Missing Items Using Naive Bayes Classifier and Graph Based Prediction

The prediction of missing items in a set is an unresolved area of research on the web. Current approaches use association rule mining techniques which are applied to only small item sets. Association rule mining techniques increase rule generation complexity as the size of data increases. This paper proposes the use of Naïve Bayes text classifier prior to the prediction process to control the transaction length thereby reducing rule generation complexity. The lengthy transactions are reduced by classification to shorter transactions, the length of which have an upper bound determined by the number of classes that are in the training dataset. The prediction of missing classes uses a graph based approach. Graph based approaches offer an advantage of low memory requirements and require just one pass over the database. The proposed approach offers advantages of prediction at a higher level of abstraction and reduced rule generation complexity.