Applied ML · AICTE Virtual Internship

Crop & Fertilizer Recommendation

An ML model recommending optimal crops and fertilizers from soil and climate data. Applied supervised learning with Python and Scikit-learn, deployed via Streamlit.

Scikit-learnPythonSQLStreamlit

Role: ML intern — AICTE Virtual Internship
Timeline: 2023

View on GitHub ↗

Field notebook UI — 'Read the soil before you plant the season.' with input sliders and 91% confidence Rice recommendation — Recommendation — 91% confidence

97.3% accuracy card across 2,200 held-out soil samples with rice, maize, cotton, wheat crop classes — Accuracy — 97.3% / 22 classes

Problem

Small farmers need data-driven crop and fertilizer choices based on their soil and local climate, not generic advice.

Approach

01Cleaned and joined soil-nutrient and climate datasets.
02Trained supervised classifiers (Random Forest, Gradient Boosting) with Scikit-learn.
03Compared models on accuracy and F1; selected the most stable across regions.
04Deployed a Streamlit app so non-technical users can input soil values and get recommendations.

Stack

Python
Scikit-learn
pandas
SQL
Streamlit

Outcomes

→Delivered an interpretable model with feature-importance explanations.
→Demoed live to internship reviewers with real soil inputs.

View on GitHub ↗