An Improved Localized Regression Approach for Bioactive Compound Extraction from Citrullus lanatus Rind via the Central Composite Design

  • O. EGUASA Benson Idahosa University
  • S. O. OSAWE Benson Idahosa University
Keywords: Central composite design, Citrullus lanatus rind, Improved local linear regression

Abstract

The rind of Citrullus lanatus (watermelon), commonly regarded as agricultural
waste, is a rich source of bioactive compounds with significant antioxidant
properties, making it a promising raw material for applications in the food,
cosmetic, and pharmaceutical industries. This study employed Response Surface
Methodology to optimize the extraction conditions for maximizing the recovery of
phytochemicals from watermelon rind. A Central Composite Design was adopted
to investigate the effects of key extraction variables, namely extraction time and
extraction temperature, on the total extraction yield. Previous studies evaluated
four conventional regression models Linear, Two-Factor Interaction, Quadratic,
and Cubic alongside the adaptive local linear regression model with adaptive
bandwidths to determine the optimal extraction yields at low, medium, and high
levels of the extraction variables. Building on this framework, the present study
incorporated an additional critical extraction factor and expanded the
experimental design to five coded levels (-α, -1, 0, 1, +α), thereby improving model
flexibility and enhancing the precision of extraction optimization. Furthermore, an
improved adaptive local linear regression model, with variable bandwidth
selection was utilized to enhance model goodness-of-fit, minimize deviations from
the zero-residual line, and improve prediction accuracy. Comparative analyses,
supported by three-dimensional response surface visualizations, demonstrated
that the proposed model performed comparably to the Quadratic Model while
consistently outperforming the Linear, Two-Factor Interaction, Cubic, and
conventional local linear regression models in predicting the extraction of bioactive
compounds from Citrullus lanatus rind. The optimization results indicated that
the proposed model improved adaptive local linear regression model and local
linear regression achieved superior process requirements under the optimal
extraction conditions. In addition, simulation studies were conducted to evaluate
the robustness of the regression models under a misspecification parameter, γ,
within the interval 0≤γ≤1. The results demonstrated that the simulated improved
adaptive local linear regression model remained stable as γ increased from 0 to 1,
confirming its robustness against model misspecification. Overall, the findings
highlight the effectiveness of the improved local linear regression approach for
optimizing bioactive compound extraction.

Author Biographies

O. EGUASA, Benson Idahosa University

Department of Physical Sciences, Benson Idahosa University, Benin City, Edo State

S. O. OSAWE, Benson Idahosa University

Department of Biological Sciences, Benson Idahosa University, Benin City, Edo State

Published
Y-m-d