FCA

Family Dynamics Analysis in France Using FCA

Project Description

This project applies Formal Concept Analysis (FCA) to analyze family dynamics in France. It identifies patterns in family structures, single-parent households, childbearing decisions, and naturalization trends across five regional zones using census data.

---

Table of Contents

1. Project Overview
2. Methodology
3. Technologies Used
4. Data Sources
5. Installation and Execution
6. Results Summary
7. Limitations
8. References

---

Project Overview

The study examines:

• Family structure and cohabitation patterns.
• Characteristics of single-parent households.
• Factors influencing childbearing decisions.
• Trends in naturalization through marriage.

By employing FCA, the project extracts meaningful patterns and generates visual representations (concept lattices) to uncover relationships within census data.

---

Methodology

1. Problem Definition: Analyze family-related variables to identify significant patterns.
2. Data Preparation:
   • Extracted a sample of 3 million rows per region from the national census dataset.
   • Features include marital status, nationality, household size, and professional status.
3. Technical Implementation:
   • Utilized Python libraries: FCApy, caspailleur, and paspailleur.
   • Data mining algorithms were used to identify stable and interesting patterns.
4. Results Analysis:
   • Generated concept lattices to visualize relationships among variables.
   • Measured pattern support and stability using Delta measures.

---

Technologies Used

• Programming Language: Python
• Libraries:
  • FCApy (Formal Concept Analysis)
  • caspailleur (Pattern mining for binary data)
  • paspailleur (Pattern mining for complex data)
• Hardware: NVIDIA A100 GPU (PNY Server)

---

Data Sources

• Data provided by The National Institute of Statistics and Economic Studies (INSEE).
• Dataset: "Housing, individuals, activity, educational and occupational mobility, residential migration in 2016."
• Regional zones analyzed:
  • Zone A: Île-de-France
  • Zone B: Centre-Val de Loire, Bourgogne-Franche-Comté, Normandie, Hauts-de-France
  • Zone C: Grand Est, Pays de la Loire, Bretagne
  • Zone D: Nouvelle-Aquitaine, Occitanie
  • Zone E: Auvergne-Rhône-Alpes, Provence-Alpes-Côte d’Azur, Corse, Overseas regions

---

Installation and Execution

Prerequisites

• Python 3.8+
• Required libraries installed:

  pip install FCApy
  pip install caspailleur
  pip install paspailleur

Steps to Run the Project

1. Clone the Repository:

   git clone <repo-link>
   cd <repo-directory>

2. Prepare the Dataset:
   • Place the dataset in the data/ directory.
3. Run the Analysis:

   python main.py

4. View Results:
   • Outputs include pattern analysis, concept lattices, and stability measures.

---

Results Summary

Key Findings:

• Family Structure: Predominantly nuclear families with 2-5 members; emerging trends in shared accommodations (Zone C).
• Single-Parent Households: Common across all zones, often with extended family support (Zones C, D, E).
• Childbearing Decisions: Influenced by marital status and employment stability; higher education may delay family formation.
• Naturalization Trends: Significant among Portuguese, Algerian, Moroccan, and Italian populations, driven by historical and geographical ties.

Visual Outputs:

• Concept lattices for each zone (Appendix).
• Scatter plots showing pattern support and Delta measures.

---

Limitations

1. Resource Constraints: High memory requirements (100GB RAM) limited dataset size.
2. Algorithmic Limitations: Extracted patterns were less precise due to implementation constraints.
3. Dataset Scope: Excluded marital subcategories (e.g., divorced, widowed) that could provide deeper insights.

---

References

1. INSEE: French National Census Data.
2. Buzmakov et al. (2023), Data Complexity: An FCA-Based Approach.
3. Uta Priss (2023), FCA Software Tools.
4. Kuznetsov & Obiedkov (2002), Formal Concept Analysis Algorithms.
5. GitHub Libraries: FCApy, caspailleur, paspailleur.