Abstract: Disclosed herein is a method of classifying financial transactions by usage patterns of a user. The method includes analyzing metadata extracted from the information associated with financial transactions in accordance with at least one business rule. Where the analysis includes sequentially analyzing the metadata using a constant-time lookup data structure, a Radix tree, a Lucene tree and fuzzy logic methods, until a unique identifier is found that is associated with the metadata. The metadata with the unique identifier is then added to the constant-time lookup data structure to update the constant-time lookup data structure. The transaction data is then classified based on the unique identifier.

Abstract: Disclosed herein is a method for providing users with saving opportunities of interest. The method includes developing a predictive model of user purchase behavior, evaluating a plurality of available savings opportunities from a savings opportunity data base using the model of user purchase behavior, selecting from the plurality of savings opportunities one with the highest likelihood of acceptance and communicating the selected savings opportunity to the user.