Soft Computing Evaluation Logic

3.7Sensitivity Analysis and Tradeoff Analysis

The process of building complex criteria includes activities of verification and validation. It is important to verify that the properties of complex criteria are consistent both with general properties of human reasoning and with goals and expectations of a specific evaluator. Two basic processes that are used for tuning and justification of complex criteria are sensitivity analysis and tradeoff analysis.

3.7.1 Sensitivity Analysis

The goal of sensitivity analysis is to investigate how various inputs and parameters affect the results of evaluation. The goal of tradeoff analysis is to investigate compensatory features of inputs of complex criteria. These analyses are auxiliary processes that contribute to the quality of decision models and the reliability of evaluation.

In a general case, let us consider a criterion function where the output y depends on n inputs and k independent parameters:

A typical structure of a general LSP criterion function is shown in Fig. 3.7.1. Inputs x₁, …, x_n can be either input attributes (a₁, …, a_n) or the corresponding attribute suitability scores (s₁, …, s_n), or subsystem suitability scores (e.g. S_im), or parameters (weights, or andness/orness). The output y can be any value affected by a selected input value. For example, the attribute a_i in Fig. 3.7.1 affects the value of subsystem suitability ...