Digital society faces the challenge of information explosion. People
urgently need new techniques and tools automatically and intelligently to extract
knowledge from vast data. So data mining has become a new hotspot. This
dissertation mainly focus on the mining of association rules in data mining, the
qualitative quantitative conversion model – cloud model is be studied and
extended. How to apply data mining into the network management is also
discussed. The main contribution of this dissertation is following:
It expounds the one-dimensional and two-dimensional cloud model,
presents the whole algorithms of various cloud generators. It studies the
mechanism of uncertainty reasoning based on cloud model and present relevant
algorithms.
It introduces the concept of association rules, provides the Apriori
algorithm to mining association rules and discusses the related study about
association rules. Then the quantitative association rules mining based concept
partition is discussed, a new algorithm is presented. At the same time, mining
weighted association rules is presented, because the classic theorem in Apriori is
no longer hold true, two algorithms are presented to solve this problem.
According to the difficult of obtaining data source, a method to generate
data source to test data mining algorithm based on cloud model is presented.
The generation algorithm for quantitative data and categorical data are given.
Because of the randomness and fuzziness in the cloud model, the generated data
contain potential knowledge except known knowledge, and the data distribution
and relationship between attributes also have random and fuzzy facts.
Finally, how to apply data mining to network management is also discussed
and depending on example, the whole procedure of data mining for alarm data
in network management is described.
Key Words: data mining, assocition rule, cloud model, cloud transform,
network management 数据挖掘研究院
