Abstract: Inadequate or restricted data, missing data, ambiguous and noisy data are all examples of characteristics that might lead to data set uncertainty. As the datasets used in this research showed, when a data set is noisy or ambiguous, the result is uncertainty. Real-world datasets are far from perfect: they are typically affected by various types of uncertainty (including missing values) that are primarily related to either the data collection technique or the complexity (e.g., volatility) of the phenomena under investigation, or both. These kinds of uncertainties are usually classified into two groups: Data that isn’t there and supervision that isn’t up to standard. In this research, goal is to handle uncertainty using various techniques while also disambiguating the uncertain situations. In this paper, a framework for uncertainty estimation and handling using several strategies is proposed. The optimal decision tree (ODT) approach is given for increasing the predictive performance of any model when the ML model is selected as a decision tree. Precision, recall, f-measure, and accuracy are some of the evaluation parameters used to evaluate the proposed ODT method. For handling uncertainty, an optimum strategy is identified based on various comparisons.

Keywords: Uncertainty, Decision Tree, Classification, Gini Impurity and Optimal Decision Tree (ODT).