🔍🌍 Econoscope: Forecasting Unemployment Rates

📑Table of Contents

• 📖Project Overview
• 🛠Installation
• 📊 Dataset Creation
• 🤖Models
• 🚀Usage & deployment
• 📁Files

📖Project Overview

Econoscope is a robust project aimed at analyzing and forecasting unemployment rates using cutting-edge machine learning techniques and time series analysis. The project leverages datasets from global organizations to provide insights into employment trends and predict future unemployment rates.

✨ Key Features:

• Accurate Forecasts: Utilizes advanced machine learning models to predict unemployment rates with high accuracy.
• Interactive Visualizations: Provides interactive charts and graphs to visualize historical and forecasted data.
• User-Friendly Interface: Easy-to-navigate interface for exploring forecasts and historical data.
• Customizable Predictions: Allows users to input custom parameters for tailored forecasts.
• Comprehensive Data: Leverages extensive datasets from global organizations for robust analysis.

🛠Installation

Clone repository:

git clone https://github.com/sofibrezden/econoscope.git

Create a Virtual Environment:

python -m venv .venv 

On Linux:

source .venv/bin/activate  

On Windows:

.venv\Scripts\activate

For backend:

Install dependencies:

pip install -r .\app\requirements.txt

Run Flask server:

flask run

For frontend:

Open a new terminal or use an existing one after activating the backend. Navigate to the frontend directory:

cd your-repo-name/frontend

Install dependencies using npm:

npm install

Start the React application:

npm start

For visualization server:

Open a new terminal or use an existing one.

Navigate to the visualization directory:

cd your-repo-name/graphs

Install the required packages from the requirements.txt file:

pip install -r requirements.txt

Start the server:

python main.py

The application will be available on http://localhost:3000. Ready to use!

📊Dataset Creation

To analyze labor market trends and unemployment rates, two datasets were utilized:

🗂ILOSTAT Dataset

Provides unemployment rates globally, categorized by gender, age groups, and years (2000-2023). Key features include:

• Value: Unemployment rate.
• Sex: Male, female, and total.
• Age: Age groups.
• Country: Country name.
• Year: Year of data.

🗂Kaggle "Global Unemployment Data" Dataset

Contains annual unemployment rates (2014-2024) for countries worldwide, categorized by demographics such as gender and age groups. Key features include:

• Country Name: Country.
• Indicator Name: Type of unemployment indicator.
• Age Categories: Demographics by age.
• Annual Rates: Unemployment rates for years 2014-2024.

Additionally, the Country-Continents Dataset was used to add a "Continent" feature, enabling analysis at both country and continental levels. The datasets were merged to provide comprehensive data on unemployment rates across genders, age groups, countries, and continents.

🧹 Handling Missing Data

• Missing values were filled using averages for each country and year.
• Linear extrapolation was applied to estimate missing data for certain countries.
• Countries with insufficient data were excluded to ensure dataset reliability.

⚖️ Balancing

• The "Sex" column was balanced by computing missing "Total" values as averages of "Male" and "Female."

📈Enhancements

• Monthly unemployment data was calculated using labor force data from ILOSTAT, improving granularity for 27 countries.
• OECD data added monthly unemployment data for 24 more countries.

The final dataset contains 1,477 rows and 16 columns, ready for advanced analysis and forecasting.

🤖Models

Econoscope employs multiple machine learning models to forecast unemployment rates:

• ARIMA: A time series model combining autoregression, integration, and moving average components. It captures short-term dependencies and long-term trends, suitable for stable trends without seasonal effects.

• SARIMA: Extends ARIMA by including seasonal components. Effective for datasets with significant seasonal patterns, such as annual labor market cycles.

• XGBoost: A powerful gradient boosting algorithm that handles nonlinear relationships and multi-factor data. Ideal for capturing complex interactions but requires careful feature engineering for time-series tasks.

Each model is evaluated using key metrics such as MSE, RMSE, and MAPE.

🚀Usage

The Econoscope website is deployed and accessible via this link: 👉 Econoscope Live Website.

• The backend and visualization server are hosted on the Render platform, while the frontend is deployed on Vercel.
• The PostgreSQL database is also hosted on Render’s cloud infrastructure.

Note: Due to the free subscription plan on Render, requests to the backend and visualization server may experience delays of up to 50 seconds. This is expected behavior under the free-tier plan and does not impact the accuracy or functionality of the application.

To explore the forecasts and insights, simply visit the website and interact with the intuitive interface.

📁Files

• app/: Backend Flask application files.

• frontend/: React-based frontend files.

• datasets/: Dataset files used for analysis and forecasting.

• graphs/: Visualization server files.

• notebooks/: Jupyter notebooks for exploratory analysis and forecasting, including:

  • creating_new_dataset&EDA.ipynb: Covers dataset creation and EDA. Combines data from multiple sources (e.g., Kaggle, ILOSTAT), adds continent info, cleans data, analyzes trends, and explores correlations between unemployment, age, sex, and region.

  • forecast_2025.ipynb : A notebook for forecasting unemployment rates for 2025. Evaluates model performance and visualizes forecasts.

  • forecast_2026.ipynb : A notebook for forecasting unemployment rates for 2026. Evaluates model performance and visualizes forecasts.

  • forecast_2027.ipynb : A notebook for forecasting unemployment rates for 2027. Evaluates model performance and visualizes forecasts.

🤝 Contributing

Contributions are welcome! Please open an issue or submit a pull request for any improvements or bug fixes.