Rule Mining Algorithm

All the organizations generate and collect huge volumes of data that they use in daily operations. The necessary data is captured and maintained by the corresponding department for each of its operations. Despite this wealth of data, many companies were unable to fully capitalize on its value because most of the information that are implicit in the data are not easy to find it out. To take advantage of high return profits and to compete effectively in the market, decision-makers must be able to find and utilize the hidden information in the collected data. Automated systems has contributed to the production of large volumes of data.

As the volume of data becomes so large,there is a need for tool to analyze it. Data mining is the automatic

This brought new challenges, greater demands, and new research directions. To discover the knowledge in the database,several efficient association rule mining algorithms are developed. But when we apply these approaches to the real time problems,we face some glitches in it. The reasons are,

1.Association rule mining algorithm generate large volume of pattern and rules and thus the process becomes a time consuming process. Due to enormous data, users cannot use and maintain the knowledge.

2.The pattern uses a subset relationship,so it is difficult to use the structural information to interpret the patterns. More number of patterns causes interpretability issue. Thus the importance of knowledge discovery reduces significantly.

3.Noise and uncertainty contains in the discovered knowledge .If the volume of the result is too large then there will be more redundancy in rules and patterns. However,there will be unnecessary patterns and rules in the result,which are not interesting for

Support and confidence

These two approaches lack some specific rule or patterns. Identifying low support and confidence improves the coverage but results in huge amount of rule. Thus some approaches have been proposed to address these problems. For frequent patterns,two approaches are used to reduce the number of patterns. The coverage of resource can be specified. Redundant items may occur during the process. The main task of rule generation is to eliminate redundant rules and unnecessary rules. So during rule generation,some form of constraints can be set. Based on the above observation,one can reduce the number of extracted rules. However, some amount of redundancy remains. Further there occurs some patterns which are not really interested by the user also remains as well. In a table, a row is called a granule and it consists of relevant attributes. With decision rules, one can reduce the two phases of association rule mining into one stage. But when we try to directly apply these decision rules,it creates problems.

1.The relationship between the patterns and granules are not well understood.

2.One cannot identify meaningless rules and accesss the rules