Rapid and accurate search similar case is the key of establishing CBR engine design system. In order to enhance the search speed, a novel mixed FCM clustering is used to establish category index of CBR engine design system. Firstly, because date type of engine general parameters includes quantitative, Boolean and categorical data, categorized concept tree is used to quantify category parameters and present the measure of mixed similarity. Secondly, because traditional fuzzy C mean (FCM) algorithm easy get into local best solution, calculate slowly and is obviously influenced by noise data, this paper combines ant colony algorithm and FCM, uses ant transition probability as initial value of membership matrix to calculate center and adopts calculated center to initialize FCM center. Experiment results show improved FCM clustering algorithm can increase search efficiency obviously. Finally, the engine design system is established based on improved FCM algorithm and is utilized during engine design process.
[1]
Cheng-Fa Tsai,et al.
ACODF: a novel data clustering approach for data mining in large databases
,
2004
.
[2]
Chen Ning.
Fuzzy K-Prototypes Algorithm for Clustering Mixed Numeric and Categorical Valued Data
,
2001
.
[3]
Jiawei Han,et al.
Knowledge Discovery in Databases: An Attribute-Oriented Approach
,
1992,
VLDB.
[4]
M. Dorigo,et al.
Ant System: An Autocatalytic Optimizing Process
,
1991
.
[5]
Zhaohao Sun,et al.
R5 model for case-based reasoning
,
2003,
Knowl. Based Syst..
[6]
Marco Dorigo,et al.
Ant system: optimization by a colony of cooperating agents
,
1996,
IEEE Trans. Syst. Man Cybern. Part B.