Logo Lanfrica

unityaaron/agricultural-input-subsidy-simulator

Domaine:

agriculturesocioeconomic

Type de record:

softwaremodel
Créateur:
uni
Hôte:
Production-grade ML and microeconomic policy simulation tool evaluating smallholder agricultural input subsidies, price shocks, and factor efficiencies in Nigeria. # 🌾 Sub-Saharan Smallholder Input Subsidy Evaluation & Profitability Simulator ### Developed by Unity Aaron An end-to-end data engineering and predictive machine learning system designed to simulate input subsidy viability, capture market price volatility, and isolate agricultural intervention causality. This system pairs a robust statistical inference engine with an optimized Random Forest Regressor to serve real-time microeconomic policy simulations for smallholder farming systems in Edo State, Nigeria. 🔗 **Live Interactive Web Application:** Click Here to Launch the Policy Simulator --- ## 📊 Core Architecture & Analytical Modules The repository is engineered systematically to separate experimental data science from live software deployment: * **`statistical_inference/`**: Houses multiple OLS regression modeling used to isolate the true treatment effects of agricultural extension training while strictly controlling for confounding physical resource factors (Farm Size, Input Volumes) to address baseline asset bias. * **`machine_learning_model/`**: Contains the predictive engines, featuring an optimized Random Forest Regressor model, cross-validation scripts, and frozen deployment assets (`.pkl`). * **`run_pipeline.py`**: An automated, single-click ETL pipeline that ingests raw micro-level operational spreadsheets, executes feature calculations, and exports live predictions. * **`app.py`**: A public-facing web interface that couples the ML yield prediction backend with real-time financial cost-benefit matrix formulas to simulate fertilizer profitability shocks. --- ## 🧠 Strategic Research Insights & Metrics ### 1. Causal Inference Verification Through independent T-tests and multiple linear regression controls, the system isolates training as a primary causal driver ($p < 0.05$), unlocking a **15.68 to 16.5 percentage point increase** in yield gains, while baseline fertilizer volume alone proved statistically insignificant under prevailing unmanaged field co …