Efficient methods for mining sequences in large data sets

OData support
Supervisor:
Nagy István
Department of Telecommunications and Media Informatics

The thesis attempts to elaborate a sequence-based mining method for predicting customer behaviour, optimized for business use. This requires the knowledge of the currently applied systems, including their construction and use. We also have to apprehend the mathematical methods in order to rectify the shortcomings of the existing systems.

During the sequence mining in the developed procedure, the events (and time intervals among the events) are represented as a fuzzy set. The sequences found during the process are represented in fuzzy relations. These sequences, and the transitional probabilities between them, are presented in a Markov-chain (with the necessary time handling), that become a predictive model.

As a result of this new method, a financial institution can analyze and forecast the behaviour of its customers or the future use of its products in one single model. Additionally, this new type of sequence mining supplies more practicable information about the behaviour of our customers.

Downloads

Please sign in to download the files of this thesis.