Evaluation of the Efficiency and Comparison of Different Numerical Differentiation Methods on Three Case Studies

TL;DR

This paper evaluates and compares the effectiveness of different numerical differentiation methods across three diverse case studies, highlighting their strengths and limitations in practical applications.

Contribution

It provides a comprehensive comparative analysis of six numerical differentiation techniques applied to real-world case studies, revealing their relative performance and suitability.

Findings

Forward and Backward methods perform well with low computational cost

Centered method yields better results in temperature variation case

None of the methods are satisfactory for market equilibrium estimation

Abstract

Without question regarding its pivotal significance, the computation of function derivatives carries substantial weight within a multitude of engineering and applied mathematical fields. These encompass optimization, the development of nonlinear control systems, and the assessment of noisy time signals, among others. In this study, we have chosen three illustrative cases: the logistic model for population dynamics, temperature variation within buildings, and the determination of market equilibrium prices. The primary objective is to assess the effectiveness of various numerical differentiation techniques and conduct a comparative analysis of the outcomes for each of these case studies. To achieve this objective, we employed three distinct numerical differentiation techniques: The Forward, Backward, and Centered FiniteDifference methods, each executed in two different levels of precision, totaling six variations. Our findings clearly indicate that, for the initial case study, the methods characterized by lower computational costs (specifically, the Forward and Backward Finite-Difference methods) yield superior outcomes. In contrast, for the second case study, the Centered Finite-Difference method delivers better results. In the case of the third case study, our results reveal that none of the methods produce estimations that meet acceptable standards. It is noteworthy that the empirical equations for each case study have been validated against previous literature.