Traditional call tree classifiers work with information whose values area unit celebrated and precise. we have a tendency to extend such classifiers to handle information with unsure info. worth uncertainty arises in several applications throughout the information assortment method. Example sources of uncertainty embrace measurement/quantization errors, information staleness, and multiple recurrent measurements. With uncertainty, the worth of a knowledge item is commonly depicted not by one single worth, however by multiple values forming a probability distribution. instead of abstracting unsure information by applied math derivatives (such as mean and median), we have a tendency to discover that the accuracy of a call tree classifier will be a lot of improved if the “complete information” of a knowledge item (taking into account the chance density operate (pdf)) is utilized. We extend classical call tree building algorithms to handle data tuples with unsure values. in depth experiments have been conducted that show that the ensuing classifiers area unit a lot of accurate than those victimization worth averages. Since process pdf’s is computationally a lot of pricey than process single values (e.g., averages), call tree construction on unsure information is more electronic equipment demanding than that sure information. To tackle this problem, we have a tendency to propose a series of pruning techniques which will greatly improve construction potency.