Genetic Algorithm-Based Feature Selection and Random Forest Modelling of Agricultural Biotechnology Adoption and Crop Yield in Tamil Nadu

DOI

https://www.doi.org/10.59256/indjcst.20260503037

Published In

Volume 5, Issue 3 (2026)

Published Online

September-December 2026

Pages

314-319

Article Type

Original Article

Check for updates

This research paper attempts to identify the Biotechnology inputs such as tissue-culture seedlings, improved seed, microbial biofertilisers and bio pesticides are promoted widely in Tamil Nadu, yet their combined association with crop performance is seldom examined across many crops at once. This study analyses 6,500 farm-level records covering 38 districts, 14 crops and the years 2021–2025 (a simulated research dataset) and introduces a relative yield index (RYI) that expresses each yield as a ratio to its crop median, so that crops with very different absolute yields can be compared. A genetic algorithm (GA) searches the space of 41 candidate predictors, using cross-validated Random Forest accuracy with a small size penalty as fitness, and a final Random Forest is trained on the selected subset. The GA retained 23 predictors (a 43.9% reduction) and reached a hold-out R² of 0.382 and RMSE of 0.140, against 0.367 and 0.141 for a Random Forest on all predictors; gradient boosting scored highest (R² = 0.418). Tissue-culture seedlings (+10.1%, Cohen d = 0.58) and improved seed (+7.2%, d = 0.40) were linked to a higher RYI, whereas farmer training showed no detectable effect. Irrigation source was the dominant predictor: drip or sprinkler farms exceeded rain fed farms by 27%.