Decision Tree Analysis Tool with the Design Approach of Probability Density Function towards Uncertain Data Classification

Abstract

Traditional decision tree classifiers are built utilizing certain or point data as it were. Be that as it may, in numerous genuine applications innately data is constantly uncertain. Quality or esteem uncertainty is characteristically connected with data esteems amid data gathering process. Traits in the preparation data sets are of two kinds – numerical (constant) and clear cut (discrete) characteristics. Data uncertainty exists in both numerical and all out characteristics. Data uncertainty in numerical qualities implies scope of qualities and data uncertainty in all out traits implies set or accumulation of qualities. In this paper we propose a technique for taking care of data uncertainty in numerical properties. One of the least difficult and most straightforward techniques for taking care of data uncertainty in numerical properties is finding the mean or normal or agent estimation of the arrangement of unique estimations of each estimation of a characteristic. With data uncertainty the estimation of a property is generally spoken to by an arrangement of qualities. Decision tree classification precision is tremendously enhanced when property estimations are spoken to by sets of esteems as opposed to one single delegate esteem. Probability density function with equal probabilities is one compelling data uncertainty demonstrating system to speak to each estimation of a property as an arrangement of qualities. Here the principle presumption is that genuine esteems gave in the preparation data sets are found the middle value of or delegate estimations of initially gathered esteems through data accumulation process. For every illustrative estimation of each numerical characteristic in the preparation data set, approximated values relating to the initially gathered esteems are created by utilizing probability density function with equal probabilities and these recently produced sets of qualities are utilized as a part of developing another decision tree classifier.