Beccaworsnop/Algerian-Market-prices-ML-spread-prediction

This project analyzes Algerian market price data to predict price relative spreads between lower and upper bounds, providing insights into market volatility and price variations across different product categories. ## Project Overview The Algerian Market Price relative Spread Prediction project implements a data science workflow to analyze and predict price relative spreads in the Algerian market. The project covers: 1. Data Cleaning & Preprocessing 2. Exploratory Data Analysis (EDA) 3. Machine Learning Model Development ## Dataset The dataset contains information about various products in the Algerian market with the following key columns: - `lowerBound`: Minimum observed price for a product - `upperBound`: Maximum observed price for a product - `price`: Average or standard price for a product - `category`: Product category (e.g., fruits, vegetables, etc.) - `month`: Month when the price data was collected - `spread`: Calculated price spread (target variable) The relative spread is defined as: `(upperBound - lowerBound) / lowerBound` and represents the relative price range for each product. ## Workflow ### 1. Data Cleaning - Calculated the relative spread between upper and lower price bounds - Handled infinity values that resulted from division operations - Removed or replaced NaN values to ensure data quality - Verified data cleanliness with statistical checks ### 2. Feature Selection & Engineering - Selected relevant features for model training: `lowerBound`, `upperBound`, `price`, `category`, and `month` - Processed categorical variables using one-hot encoding ### 3. Model Development We implemented a **Random Forest Regressor** to predict price relative spreads with the following specifications: - 100 decision trees (`n_estimators=100`) - Standard train-test split (80% training, 20% testing) - Categorical features encoded using pandas `get_dummies` functionality ### 4. Model Evaluation The model was evaluated using multiple metrics: - Root Mean …