Predictive Model for Shea Stearin Applicability Using Refractive Index

Project Abstract

This project aimed to replace expensive and time-consuming wet-chemistry analysis with a rapid, chemical-free proxy method for screening shea stearin quality. By establishing a correlation between Refractive Index (RI) and key quality parameters, we developed a tool to accelerate decision-making in production environments.

Technical Methodology

The core of the solution involved building and validating a pair of linear regression models to predict Iodine Value (IV) and StOSt triglyceride content directly from Refractive Index (RI) measurements.

• Data Collection & Processing: Curated a dataset of historical quality parameters and corresponding RI measurements. Implemented a robust outlier-handling strategy (e.g., Z-score, IQR) to clean the training data.
• Model Development: Developed simple linear regression models (OLS) determining the coefficient and intercept for the prediction equation:
  Y = β₀ + β₁X + ε
• Feature Assessment: validatated the assumption of linearity and homoscedasticity of residuals to ensure model reliability.

Model Performance & Validation

The models demonstrated strong predictive power, sufficient for rapid screening purposes:

Operational Impact

• Cost Reduction: Eliminated the need for solvents and chemical reagents for routine screening, significantly lowering analytical costs.
• Efficiency: Reduced turnaround time from hours (GC/titration) to minutes (RI measurement), enabling faster feedback loops for process control.
• Sustainability: Validated RI as a "green" analytical proxy, supporting laboratory waste reduction goals.