Decision Tree for Data Uncertainty with Pruning

Decision Tree is a widely used data classi�cation technique. This paper proposes a decision tree based classi�cation method on uncertain data. Data uncertainty is common in emerging applications, such as sensor networks, moving object databases, medical and biological bases. Data uncertainty can be caused by various factors including measure- ments precision limitation, outdated sour ces, sensor errors, and network latency and transmission problems. In this paper, we enhance the traditional decision tree algorithms and extend measures, including entropy and information gain, considering the uncertain data interval and probability distribution function. Our algorithm can handle both certain and uncertain datasets. The experiments demonstrate the utility and robust ness of the proposed algorithm as well as its satisfactory prediction accuracy. We propose a series of pruning techniques that can greatly improve the efficiency of the construction of decision trees.