الفهرس 
BACKGROUND AND RELATED WORKOnly 14 pages are availabe for public view
 |  | from | 16 |  |  |
| | | from | 16 | | |
Abstract
Association rules discovery is one of the important research areas in the data mining field. It is interested in finding significant relationships between items in any transactional database with respect to given threshold. With the massive amount of data flow every day, these transactional databases are updated with new transactions that make any prior knowledge invalid. The trivial solution is to run the mining algorithms again without utilizing the previously discovered knowledge. This would be a nightmare process especially with the large database sizes nowadays. The need of new efficient algorithms to deal with the incremental mining of association rules becomes very critical.
In this thesis, the researcher is interested in developing a set of efficient algorithms for mining interesting association rules in large incremental databases mainly with two types of databases; traditional and temporal databases. Recently, there is a new interest in weighted databases. Weights give each item another value that defines its significance in the database rather than the traditional support measure. The researcher presented a new algorithm for mining frequent weighted itemsets as a start point in this hot research topic.
Many experiments were conducted using both real and synthetic large datasets to compare the performance of the proposed algorithms with the recently cited algorithms in each topic. The experiments measure the performance with respect to many aspects such as running time, memory usage analysis with different database characteristics. The results show that the proposed algorithms have great performance optimization over other
V
algorithms. In addition, the results show that the proposed algorithms achieve linear scalability when dealing with large databases even at very low minimum support thresholds.
Keywords: Data Mining, Association Rules Discovery, Large Databases, Temporal Databases, Frequent Weighted Itemsets, Trie Indexing.