/Project-Predictive-Modeling-for-Agriculture

Harness the power of machine learning to help farmers optimize their crop selection. This project predicts the best crop type based on soil metrics like nitrogen, phosphorous, potassium levels, and pH value. Using logistic regression models, we identify the most predictive soil feature to maximize crop yield efficiently.

Primary LanguageJupyter Notebook

Sowing Success: Machine Learning for Crop Selection

Overview

This project aims to assist farmers in selecting the best crop to plant based on soil conditions. By leveraging machine learning techniques, we predict the optimal crop type using soil metrics such as nitrogen, phosphorous, potassium levels, and pH values. The goal is to maximize crop yield while considering soil suitability and agricultural efficiency.

Dataset

The dataset soil_measures.csv provided contains:

  • N: Nitrogen content ratio in the soil
  • P: Phosphorous content ratio in the soil
  • K: Potassium content ratio in the soil
  • pH: pH value of the soil
  • crop: Categorical values representing various crop types (target variable)

Each row in the dataset represents soil measurements for a specific field, with the corresponding optimal crop specified in the crop column.

Project Structure

  • soil_measures.csv: Dataset containing soil metrics and optimal crop types.
  • crop_selection.ipynb: Jupyter notebook containing the project code for data exploration, model training, and evaluation.
  • README.md: This file providing an overview of the project.

Methodology

1. Data Exploration and Preparation

  • Load and inspect the dataset.
  • Separate features (N, P, K, pH) from the target variable (crop).

2. Model Building

  • Split the data into training and test sets.
  • Build logistic regression models for each soil metric (N, P, K, pH) to predict crop types.
  • Evaluate model performance using the F1-score metric.

3. Identifying the Best Predictive Feature

  • Determine the most important soil metric for predicting crop types based on F1-score.

Results

  • The model identified K (Potassium content) as the most predictive feature for determining crop types, achieving an F1-score of 0.239.

Usage

  1. Clone the repository: 
    git clone https://github.com/NonsoOmoko/Project-Predictive-Modeling-for-Agriculture.git
cd Project-Predictive-Modeling-for-Agriculture