MohammedBardaoui/Time-Series-Modeling-of-Moroccan-Unemployment-Rates

A data-science project focused on modeling the unemployment rate in Morocco using advanced time-series techniques (ARIMA, SARIMA...etc). Includes model diagnostics, forecasting, and reproducible code. ## 📘 Project Overview This project focuses on analyzing and forecasting the **unemployment rate in Morocco** using advanced time series modeling techniques. The study uses official quarterly data published by the **High Commission for Planning (HCP)** and spans the period **2006–2025**. The goal is to understand unemployment trends across key socio-economic dimensions: - Urban vs Rural areas - Education levels - Gender - Age groups - National total (“Ensemble”) Several statistical and deep learning models were implemented to assess forecasting performance. --- ## Objectives - Analyze historical unemployment trends in Morocco. - Identify disparities between population groups. - Build forecasting models: **ARIMA, SARIMA, Holt-Winters, RNN, LSTM**. - Compare the accuracy and performance of each model. - Develop a **web application** for interactive visualization and forecasting. --- ## Dataset The dataset comes from **HCP’s quarterly employment surveys** and includes: - **Period:** 2006 Q1 → 2025 Q4 - **79 observations** - **Features include:** - Area: Urban, Rural, National - Education levels: No diploma, Medium, Higher - Gender: Male, Female - Age groups: 15–24, 25–34, 35–44, 45+ The dataset contains **no missing values** and required only minimal preprocessing. --- ## Methodology ### 1. Exploratory Data Analysis (EDA) - Visualization of time series by category - 4-quarter moving averages - Trend and seasonal pattern detection - Stationarity tests (ADF) ### 2. Forecasting Models The following models were implemented: | Model | Type | Strengths | |-------|------|-----------| | **ARIMA** | Statistical | Good for differenced series | | **SARIMA** | Statistical | Handles quarterly seasonality | | **Holt-Winters** | Exponential smoothing | Trend + seasonality | | **RNN** | Deep Learning | Learns time dependencies | | **LSTM** | Deep Learning | Captures long-term patterns | --- ## Results & Model Comparison ### Key Conclusion > **SARIMA achieved the best f …