Mining System, Based On The Constraint Concept Lattice Stellar Spectral Data Classification Rules

Concept lattice, with characteristics of completeness and accuracy, is an effective tool for data analysis and knowledge discovery. Taking customer's interest about data set as background knowledge, and guiding the process of constructing concept lattice, a new concept lattice (constrained concept lattice) is presented, so that the utility and pertinence of concept lattice construction are improved. In this paper, the incremental construction algorithm and classification rule acquisition algorithm based on constrained concept lattice are studied by taking LAMOST as background. The main research works can be summarized as follows:First, an incremental construction algorithm of constrained concept lattice based on pruning is presented. By making use of the rigorous monotone relation between father concept's intent and child concept's intent, the redundant information in the construction process is eliminated by using pruning technology, namely, the comparative operations between the intents are decreased, so that the efficiency of constructing the constrained concept lattice is improved. The experiment results validate the correctness and validity of the algorithm by taking the celestial spectrum data as the formal context.Second, a classification rule acquisition algorithm based on constrained concept lattice is presented. Firstly, the attributes in data set are divided into condition attributes and classification attributes. According to the different values of classification attribute, the data set is divided into equivalence partition Gi(1≤i≤m), and the constrained concept lattice is construted by using the condition attributes. Secondly,scanning all nodes of the constrained concept lattice from top to down by using the concept of extent support and partition support, classification rules are mined. Finally, the experiment results validate the algorithm has the higher classification correctness by taking the UCI data sets and star spectrum data as the formal contexts. Third, on the basis of above, the mining system of classification rules for star spectra data based on constrained concept lattice are designed and realized by using VC++ 6.0 and Oracle 9i as development tools. The running results show that the classification rules mined by the system are feasible and valuable for auto-classification of massive star spectra data.