<p>Accurate sugar yield prediction from sugarcane is essential for effective production planning and reducing losses in the jaggery industry. Traditional approaches often rely on complex, multi-parameter monitoring systems that are costly and inaccessible to small-scale producers. This study systematically evaluated the hypothesis that minimal chemical composition analysis can achieve comparable accuracy to comprehensive monitoring methods. We used 2001 samples from the Erode region of Tamil Nadu, India, to develop a progressive three-model validation framework to test the Chemical Integration Hypothesis. The breakthrough Model 3 (Dual-Parameter Chemical Integration, using only the Brix value and water content) achieved <InlineEquation ID="IEq1"> <EquationSource Format="TEX">$R^2 = 96.23\%$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msup> <mi>R</mi> <mn>2</mn> </msup> <mo>=</mo> <mn>96.23</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation> while reducing feature requirements by 81.8%. Regional validation across 18 locations further demonstrated robust spatial generalizability (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">$R^2 = 94.80\%$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msup> <mi>R</mi> <mn>2</mn> </msup> <mo>=</mo> <mn>94.80</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation>, <InlineEquation ID="IEq3"> <EquationSource Format="TEX">$\hbox {SD} = 0.0221$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mtext>SD</mtext> <mo>=</mo> <mn>0.0221</mn> </mrow> </math></EquationSource> </InlineEquation>). These findings demonstrate potential for addressing generalizability challenges in sugarcane agriculture, as the dual-parameter approach maintained consistent performance across diverse locations within the study region without location-specific recalibration. This research suggests that minimal-parameter chemical analysis enables accurate, cost-effective prediction of sugar yield, with rapid deployment in resource-constrained settings, and could potentially revolutionize precision sugarcane agriculture worldwide.</p>

Enhanced Sugar Yield Prediction Using Dual-Parameter Chemical Analysis and Ensemble Machine Learning

期刊论文
发表日期：

2025年10月31日

查看全文

Kathirvel Narayanasamy,
Ilayaraja Venkatachalam

摘要

Accurate sugar yield prediction from sugarcane is essential for effective production planning and reducing losses in the jaggery industry. Traditional approaches often rely on complex, multi-parameter monitoring systems that are costly and inaccessible to small-scale producers. This study systematically evaluated the hypothesis that minimal chemical composition analysis can achieve comparable accuracy to comprehensive monitoring methods. We used 2001 samples from the Erode region of Tamil Nadu, India, to develop a progressive three-model validation framework to test the Chemical Integration Hypothesis. The breakthrough Model 3 (Dual-Parameter Chemical Integration, using only the Brix value and water content) achieved $R^2 = 96.23\%$ $R^{2} = 96.23 %$ while reducing feature requirements by 81.8%. Regional validation across 18 locations further demonstrated robust spatial generalizability ( $R^2 = 94.80\%$ $R^{2} = 94.80 %$ , $\hbox {SD} = 0.0221$ $SD = 0.0221$ ). These findings demonstrate potential for addressing generalizability challenges in sugarcane agriculture, as the dual-parameter approach maintained consistent performance across diverse locations within the study region without location-specific recalibration. This research suggests that minimal-parameter chemical analysis enables accurate, cost-effective prediction of sugar yield, with rapid deployment in resource-constrained settings, and could potentially revolutionize precision sugarcane agriculture worldwide.