AgriTrend_simulation is a research-oriented simulation project designed to study long-term crop yield behavior under environmental and management stress.
Instead of focusing on short-term prediction, the project emphasizes:
- explainability over black-box accuracy
- long-term trend analysis
- scenario-based reasoning
- automated, reproducible reporting
The system runs a complete analysis pipeline and automatically generates visualizations and a structured PDF report that summarizes:
- historical trends
- factor influence on yield
- baseline futures
- controlled intervention scenarios
This repository represents the Round 2 continuation of the project, originally developed for Hack The Winter – Round 1.
| Round | Status | Rank |
|---|---|---|
| Round 1 | Selected | 139 |
| Round 2 | Disqualified | - |
| Round 3 | - | - |
Modern agriculture is under increasing pressure from:
- climate variability and rising temperatures
- soil degradation
- growing dependency on inputs such as irrigation and fertilizer
In many observed systems:
- inputs increase every year
- yields do not improve at the same rate
- yield becomes more unstable over time
Most existing solutions either:
- focus on short-term prediction, or
- rely on opaque, API-driven, black-box models
There is a need for a transparent system that answers a simpler but more fundamental question:
If current trends continue, how might crop yield behave in the long run — and which changes actually matter?
- AI / Machine Learning
AgriTrend_simulation addresses this problem through a structured and explainable pipeline:
- Generate statistically realistic synthetic agricultural data
- Analyze long-term relationships between factors and yield
- Quantify the relative importance of yield drivers
- Project a baseline future assuming no intervention
- Compare it against controlled scenario-based changes
The system does not aim to predict exact future yields.
It is designed to reveal patterns, risks, sensitivities, and limits under sustained conditions.
Round 1 focused on establishing a clean, interpretable proof of concept.
Scope in Round 1:
- One synthetic region
- One synthetic crop
- ~25 years of tabular data
- Linear regression models (raw + standardized)
- Single best-case intervention scenario
Key insights from Round 1:
- Yield shows increasing instability over time
- Rising inputs do not guarantee yield growth
- A small number of factors dominate yield behavior
- Baseline futures carry increasing long-term risk
- Coordinated 1% improvements can stabilize trends under certain conditions
Round 1 code is preserved and accessible:
- Git tag:
round-1-final - Branch:
round-1 - Round 1 Tag:
https://github.com/Abhinav08bhatt/AgriTrend_simulation/tree/round-1-final - Round 1 Branch:
https://github.com/Abhinav08bhatt/AgriTrend_simulation/tree/round-1
Round 2 extends the prototype into a more flexible and presentable system:
- Multiple selectable synthetic crops (rice, wheat, maize, universal)
- Multiple selectable scenarios (best-case, drought stress, fertilizer optimization, climate shift)
- Dataset-aware automated PDF reports
- Cleaner project structure and documentation
- Improved interpretability and safer reporting language
Round 2 code is preserved and accessible:
- Git tag:
round-2-final - Branch:
round-2 - Round 2 Tag:
https://github.com/Abhinav08bhatt/AgriTrend_simulation/tree/round-2-final - Round 2 Branch:
https://github.com/Abhinav08bhatt/AgriTrend_simulation/tree/round-2
What the system DOES
- Generates realistic synthetic agricultural datasets
- Analyzes historical yield trends
- Quantifies factor influence using interpretable models
- Projects baseline futures
- Simulates controlled intervention scenarios
- Automatically generates graphs and a final PDF report
What the system DOES NOT do
- Predict real-world yields
- Claim causal certainty
- Use black-box or deep learning models
- Replace domain expertise
- Provide a production-grade decision system
-
System flow & architecture:
RepoWorking.md -
Step-by-step reasoning notebook:
Documentation Notebook
The notebook provides the best insight into design decisions, assumptions, and interpretation logic.
demo_video.mp4
python --versiongit clone https://github.com/Abhinav08bhatt/AgriTrend_simulation.git
cd AgriTrend_simulation- Windows
python -m venv venv- Linux/macOS
source venv/bin/activate pip install -r requirements.txtpython AgriTrend_simulation.pyReports are saved automatically in:
outputs/reports/- Synthetic data only
- Linear modeling by design
- No economic or policy constraints
- No regional calibration
These limitations are intentional to preserve clarity and explainability.
Planned extensions include:
- Interactive GUI or local app interface
- Expanded crop and regional modeling
- Improved data validation
- Integration of real-world datasets (where feasible)
- Interactive scenario exploration
This project was developed by The Eskimos! as part of Hack The Winter.
Check the commit graph for team information.
-
Abhinav
System design, data modeling, analysis pipeline implementation, scenario logic, visualization, report generation, git management. -
Anubhav
-
Hariom Chamoli
-
Abhinav Benjwal
This project is intended for academic, educational, and research demonstration purposes.