C4.5: programs for machine learning
C4.5: programs for machine learning
Mining quantitative association rules in large relational tables
SIGMOD '96 Proceedings of the 1996 ACM SIGMOD international conference on Management of data
Mining optimized association rules for numeric attributes
PODS '96 Proceedings of the fifteenth ACM SIGACT-SIGMOD-SIGART symposium on Principles of database systems
Attribute-oriented induction in data mining
Advances in knowledge discovery and data mining
Fast Algorithms for Mining Association Rules in Large Databases
VLDB '94 Proceedings of the 20th International Conference on Very Large Data Bases
Efficient Construction of Regression Trees with Range and Region Splitting
VLDB '97 Proceedings of the 23rd International Conference on Very Large Data Bases
Detecting a Compact Decision Tree Based on an Appropriate Abstraction
IDEAL '00 Proceedings of the Second International Conference on Intelligent Data Engineering and Automated Learning, Data Mining, Financial Engineering, and Intelligent Agents
Constructing appropriate data abstractions for mining classification knowledge
INAP'01 Proceedings of the Applications of prolog 14th international conference on Web knowledge management and decision support
Hi-index | 0.00 |
In this paper, we investigate data abstractions for mining association rules with numerical conditions and boolean consequents as a target class. The act of our abstraction corresponds to joining some consecutive primitive intervals of a numerical attribute. If the interclass variance for two adjacent intervals is less than a given admissible upper-bound 驴, then they are combined together into an extended interval. Intuitively speaking, a low value of the variance means that the two intervals can provide almost the same posterior class distributions. This implies few properties or characteristics about the class would be lost by combining such intervals together. We discuss a bottom-up method for finding maximally extended intervals, called maximal appropriate abstraction. Based on such an abstraction, we can reduce the number of extracted rules, still preserving almost the same quality of the rules extracted without abstractions. The usefulness of our abstraction method is shown by preliminary experimental results.