Distributed Based On Decision Tree Classification Algorithm

We are now living in a great new era of network. Communications, computer and network are changing our human beings and society. Nowadays, most classification algorithms cater to a centralized environment .As most data sets, users and modern organizations are geographically distributed, and merging data sets from different sites into a centralized site will incur huge network communication costs. Therefore it is necessary to make good use of geographically distributed data sets and combine different classification technologies for implementing high-performance distributed knowledge discovery systems.Although distributed classification algorithms is a new research domain, a lot of people focus their mind on the research and get a lot of academic success. Jerzy et al give their research on distributed classification algorithm and get a lot of good results. Agents, in a collaborative fashion, generate partial trees and communicate the temporary results to the Mediator. Distributed classification is accomplished via a synchronized collaboration of Agents as well as a Mediator component. The Mediator component facilitated the communication among Agents. The process is terminated when a final tree is induced.There are two performance evaluations to a decision tree: concision and precision. In the distributed classification algorithm based on decision tree, the important problem is the task of acquiring an integrated decision model from distributed data sets. There are two main strategies in the integration of local decision tree. One is integrating the local rules to establish final results after the local decision tree is made. The other is integrating the local decision trees to get the whole decision tree after the local decision trees are made, then the final results is got.It was proved that the second technique could acquire a better final result. In our experiment, the integrated decision model from the local decision tree shows high performance.