An algorithm for classification in data mining based on classification codes

Classification is a key data mining technique whereby database tuples acting as training samples are analyzed in order to produce a model of the given data. A number of classification techniques from the statistics and machine learning have been proposed. A well accepted method of classification is the induction of decision trees. The efficiency of existing decision tree algorithms has been established for small data sets. in decision tree, more number of rules are to be generated to classify the given data, because the algorithm performs the testing on attribute by attribute at level by level which is a time consuming and occupies more memory to store. In rule based classification, all combination of the fields in the table is to be taken to generate more rules for classifying the given data. To overcome these, a new algorithm is proposed which modifies the consideration of the decision tree for classification at the data warehousing level by grouping the samples using classification codes in each branch of the tree. At run time, only the code field and class field are transferred to main memory, which makes the effective usage of main memory, though the database is very large. With this construction, the number of rules to be generated is decreased and the number of tests to be performed also decreased which makes execution fast and increases the throughput. The proposed algorithm proves to be effective and efficient.