MARC: Mining Association Rules from datasets by using Clustering models
This addresses the problem of time-consuming and excessive rule generation in data mining for analysts, though it appears incremental as it builds on existing clustering and neural network techniques.
The paper tackles the inefficiency of symbolic models in extracting association rules from large datasets by proposing MARC, a method that uses a multi-topographic unsupervised neural network and clustering quality measures to identify important rules at two levels, reducing time and rule count.
Association rules are useful to discover relationships, which are mostly hidden, between the different items in large datasets. Symbolic models are the principal tools to extract association rules. This basic technique is time-consuming, and it generates a big number of associated rules. To overcome this drawback, we suggest a new method, called MARC, to extract the more important association rules of two important levels: Type I, and Type II. This approach relies on a multi-topographic unsupervised neural network model as well as clustering quality measures that evaluate the success of a given numerical classification model to behave as a natural symbolic model.