Machine Learning-Driven Predictive Modelling for Laboratory Evaluation of the Toxicological Efficacy of Cymbopogon citratus Ethanolic Extract Against Acanthoscelides obtectus
Post-harvest storage losses inflicted by the bean weevil (Acanthoscelides obtectus) threaten agricultural economic stability and legume food security across Nigeria. While conventional pest management depends extensively on synthetic chemical treatments, issues surrounding environmental resilience, applicator health, and toxic residue deposition require eco-friendly botanical alternatives. This study evaluates the laboratory contact toxicity of an ethanolic leaf extract of Cymbopogon citratus (Lemongrass) against adult A. obtectus and introduces a machine learning framework to model time-dose-mortality dynamics. This is an advancement in smart farming that outperforms conventional linear calculations, especially when predicting pest control outcomes. Fresh C. citratus leaves were collected, air-dried, and subjected to cold maceration extraction using absolute ethanol. Gas Chromatography-Mass Spectrometry (GC-MS) analysis of the crude extract revealed a rich matrix of bioactive constituents, prominently dominated by citral isomers (geranial and neral), myrcene, and geraniol. In controlled laboratory bioassays, adult bean weevils were exposed via filter-paper impregnation to graded volumetric doses of the ethanolic extract (0.0 to 10.0 ml). Empirical observations demonstrated strong, rapid contact toxicity, with a 10 ml dose inducing a 50% population mortality rate within 10 minutes, accelerating to complete population knockdown (100%) within 30 minutes, while untreated control cohorts experienced zero mortality. To expand these empirical thresholds into a predictive tool, an analytical dataset (n = 1,200) mapping extract dose, exposure intervals, temperature (25–30 °C), and relative humidity (60–75%) was evaluated using a Random Forest Regressor and Support Vector Regression (SVR). The Random Forest architecture yielded superior predictive performance (R2 = 0.988, MAE = 1.32%), isolating an optimal operational application benchmark of 6.8 ml to achieve near-total eradication (98.5%) within 45 minutes while minimizing active extract waste. This integration of botanical toxicology and ensemble learning provides a standardized, scalable framework for modern smart-storage pest management.
Authors
- R. T
- O. F
- Olotuah
Institutions
- Adekunle Ajasin University (NG)
Publication Details
- Journal
- Iconic Research and Engineering Journals
- Published
- 2026-11-13
- DOI
- https://doi.org/10.64388/ire1725006
- Primary Topic
- Insect Pest Control Strategies
- Type
- article
- Field-Weighted Citation Impact
- 0.00