Strawberry Price Prediction

This project aims to predict strawberry prices at a 2-week horizon using historical data and weather information. The focus is on demonstrating a structured approach to time series forecasting, from data exploration to model evaluation.

Data Description

The dataset combines weekly strawberry market prices with meteorological data:

• Temporal range: Multiple years of weekly data
• Features: Weather conditions (temperature, precipitation, etc.)
• Target: Weekly average strawberry prices
• Notable characteristic: Missing summer period data (weeks 24-49)

Technical Approach

1. Data Processing

   • No Handling of missing values, removing them
   • Implemented data scaling

2. Model Development

   • Baseline: Simple average-based prediction
   • Time Series: ARIMA modeling
   • Machine Learning: XGBoost with feature engineering

3. Evaluation

   • RMSE as primary metric
   • Visual comparison of predictions

Project Structure

The project is organized into four main notebooks, each focusing on a specific aspect of the analysis:

1. Exploratory Data Analysis (1_explore.ipynb)

• Data overview and missing values analysis
• Price distribution
• Seasonal patterns identification
• Key finding: Strong seasonal patterns and missing summer data

2. Feature Engineering (2_feature_eng.ipynb)

• Data cleaning and preprocessing
• Feature creation and processing
• Data scaling and export

3. Model Implementation (3_model.ipynb)

• Implementation of three different approaches:
  • Naive forecast (baseline)
  • ARIMA model (statistical approach)
  • XGBoost (machine learning approach)
• Model training and prediction generation

4. Model Evaluation (4_evaluation.ipynb)

• Comprehensive model comparison
• Performance metrics analysis
• Visualization of predictions

Project Organization

├── data/
│   ├── raw/           # Original dataset
│   ├── processed/     # Cleaned and processed data
│   └── interim/       # Intermediate data
├── models/            # Saved model files
└── src/
    ├── fct_eda.py           # EDA functions
    ├── fct_feature_eng.py   # Feature engineering functions
    ├── fct_model.py         # Model implementation
    ├── fct_evaluation.py    # Evaluation functions
    └── parameter.py         # Project parameters

🚀 Getting Started (Poetry)

1. Install Poetry (if not already installed)
   👉 https://python-poetry.org/docs/#installation

2. Install dependencies:

poetry install

3. (Optional but recommended) Launch a shell with the Poetry virtual environment:

poetry shell

4. Run the notebooks in the following order:

• 1_explore.ipynb

• 2_feature_eng.ipynb

• 3_model.ipynb

• 4_evaluation.ipynb