Sales Analysis and Prediction using Random Forest

This project is designed to extract sales data from a PostgreSQL database, process it, and use a Random Forest model to predict sales quantities. It also visualizes real and predicted sales for better understanding.

Project Structure

project/
|-- connection.py
|-- query.py
|-- random_forest.py

1. connection.py

This script establishes a connection to the PostgreSQL database using the psycopg2 library. It handles errors gracefully and ensures a reliable connection.

• Key Functions:
  • connection(): Connects to the database and returns the connection object.

2. query.py

This script queries the sales data from the database. It groups sales by year, month, and product name and calculates the total quantity sold and revenue for each group.

• Key Functions:
  • query(): Executes the SQL query and returns the grouped sales data.

3. random_forest.py

This script uses the queried sales data to train a Random Forest Regressor model and predicts future sales. It also visualizes the real and predicted sales for comparison.

• Workflow:
  1. Fetches data using query().
  2. Processes and transforms the data.
  3. Trains a Random Forest Regressor on historical sales data.
  4. Evaluates the model using Mean Squared Error (MSE).
  5. Visualizes real and predicted sales.

Prerequisites

• Python 3.8+
• PostgreSQL database with a sales table.

Required Python Libraries

Install the dependencies using pip:

pip install psycopg2 pandas matplotlib seaborn scikit-learn

Database Table Structure

The project assumes a sales table with the following structure:

Column	Type	Description
date	DATE	Date of the sale
product_name	TEXT	Name of the product
quantity_sold	INTEGER	Quantity of the product sold
unit_value	FLOAT	Total sales value of the product

How to Run the Project

1. Set up the PostgreSQL Database:

   • Ensure the sales table is created and populated.

2. Configure the Database Connection:

   • Update the connection.py file with your PostgreSQL credentials:

     conn = psycopg2.connect(
         host="<your_host>",
         database="<your_database>",
         user="<your_user>",
         password="<your_password>",
         port="<your_port>"
     )

3. Run the Scripts:

   • Execute the scripts in the following order:

     python random_forest.py

Example Output

• Mean Squared Error: Displays the MSE of the predictions.
• Visualization: A plot comparing real sales vs. predicted sales over time.

Notes

• Ensure the locale settings in query.py (es_ES.UTF-8) match your system's configuration. If not, modify it to a suitable locale.
• This project focuses on predicting sales quantities. Future improvements could include:
  • Predicting revenue.
  • Adding seasonality and external factors to the model.
  • Using advanced machine learning techniques.