Neural Networks for Data Mining: Constrains and Open Problems
Răzvan Andonie, Boris Kovalerchuk · 2004
Abstract. When we talk about using neural networks for data mining we have in mind the original data mining scope and challenge. How did neural networks meet this challenge? Can we run neural networks on a dataset with gigabytes of data and millions of records? Can we provide explanations of discovered patterns? How useful that patterns are? How to distinguish useful, interesting patterns automatically? We aim to summarize here the state-of-the-art of the principles beyond using neural models in data mining. 1 What is special in data mining applications? Data mining (DM) is the nontrivial extraction of implicit, previously unknown, interesting, and potentially useful information (usually in the form of knowledge patterns or models) from data. Historically data mining has grown from large business database applications, such as finding patterns in customer purchasing activities from transactions databases. Original DM problems were to adjust known methods such as decision trees and neural networks (NN) to large datasets (100,000 and more records) and relational database structures. Later methods such as association rules were developed specifically motivated by DM challenge. The most vehiculated DM problems are reduced to traditional statistical and machine leaning methods: classification, prediction, association rule extraction, and sequence detection. The techniques used in DM are very heterogeneous: statistical methods, case-based reasoning, NN, decision trees, rule induction, Bayesian networks, fuzzy sets, rough sets, genetic algorithms/evolutionary programming. The following are the major stages in solving a DM problem [7]: 1. Define the problem. 2. Collect and select data, such as deciding which data to collect and how to collect them.