Knowledge Fusion Technique Using Classifier Ensemble by Combining the Sets of Classification Rules

The task of data fusion is to identify the true values of data items (e. g. , the true date of birth for Shivaji Maharaj) among multiple observed values drawn from different sources (e. g. , Web sites) of varying reliability. The task of extracting knowledge from sample data is divided into a number of subtasks in case of machine learning applications. At some point, there is necessity to fuse or to combine the knowledge. And this knowledge is now contained in a number of classifiers in order to apply it to new data. It is impossible to exchange the raw data because of a limited communication bandwidth. Also, a central unit would constitute a single point of failure. In other data mining applications, knowledge extraction is split into subtasks due to memory or runtime limitations. Again, locally extracted knowledge must be consolidated later and quite often, the communication overhead should be low. Extracting information from multiple, possibly conicting, data sources, and reconciling the values so the true values can be stored in a central data repository, is a problem of vital importance to the database and knowledge management communities.