Decision Tree Modeling Using Integrated Multilevel Stochastic Networks

Abstract

Decision trees (DTs) have proven to be valuable tools for decision making. The common approach for using DTs is calculating the expected value (EV) based on single-number estimates, but the single-number EV method has limited the DTs' real-life applications to a narrow scope of decision problems. This paper introduces the stochastic multilevel decision tree (MLDT) modeling approach, which is useful for analyzing decision problems characterized by uncertainty and complexity. The MLDT's advantages are shown through a computer simulation program: the Decision Support Simulation System (DSSS). The DSSS allows users to model probabilistic linear graph networks and provides a hierarchical modeling method for modeling decision trees to present uncertainties more accurately. It consists of three modules: tree analysis networks (TANs), the shortest and longest path dynamic programming analysis network, and cost time analysis networks. The paper only discusses the TAN module by presenting the MLDT concept under the TAN of the DSSS computer application. The content of the paper includes the modeling approach, its advantages, and examples that can be used in modeling stochastic trees. The DT-DSSS was verified by conducting several tests and validated by using it extensively for undergraduate courses in civil engineering at the University of Calgary for the last two academic years.