The Research On The Algorithms Of Mining Association Rules

Knowledge discovery in databases (KDD) is a multidisciplinary field, drawing work from areas including database technology, artificial intelligence, machine learning, statistics, neural networks, and pattern recognition. Data mining is a key step of KDD, often regarded as the synonym of KDD. It is the process of mining the interesting, potentially useful, valid and understandable knowledge in data.Association rule mining is an important sub-branch of data mining, which mines interesting association or correlation relationships among a large set of data items. Association rules are considered interesting if they satisfy both a minimum support threshold and a minimum confidence threshold. Association rule mining has become a hot research topic in recent years, and it has been used widely in selective marketing, decision analysis and business management. Association rule mining algorithms are the core contents in the area, and there are several famous typical algorithms. This dissertation does some research on these algorithms, proposes a new algorithm and applies it to distributed datamining.There are original main ideas in the paper as follows:1. The dissertation extends the concept of tranditional asssociation rule mining, and introduces some concepts, such as the length of transaction, the support count of transation, the in relation between transactions, frequent transaction and frequent sub item sets.2. CSR (compress_scan_reason) is a new association rule mining algorithm based on transaction. It scans the original database only once, compresses the database by the length of transaction and scans the compressed database twice. Compared with Apriori, CSR decreases the counts of scanning and increases the effiency of mining.3. CR is an improved algorithm of CSR. It has more efficiency, for it has a better structure in the compressed database, and reduces the scale of frequent item sets.4. The dissertation integrates CR algorithm and agent technology, andproposes a framework of multi-agent association rules mining system (MARMS), which has simple structure and is esay to be realized. It can accomplish the task of distributed association rule mining, for it has the advantadges of both mobile agent and multi-agent system.