📘 Data-Science-Community-Collaboration-Analysis

(Phân tích cộng đồng nghiên cứu & phát hiện nhóm hợp tác trong lĩnh vực Khoa học Dữ liệu)

---

📌 Overview

(Tổng quan)
This project analyzes research collaboration patterns in the Data Science field by building a large-scale co-author network using data collected from OpenAlex. Community detection algorithms (Louvain, Leiden, Fast Greedy) are applied to uncover research groups and influential authors.
(Dự án phân tích mô hình hợp tác nghiên cứu trong lĩnh vực Khoa học Dữ liệu bằng cách xây dựng mạng đồng tác giả từ OpenAlex và áp dụng các thuật toán phát hiện cộng đồng.)

---

✨ Objectives

(Mục tiêu)

• Build a co-author network representing scientific collaborations.
• Apply community detection algorithms (Louvain, Leiden, Fast Greedy).
• Compute centrality metrics (Degree, Betweenness, Closeness, Eigenvector).
• Visualize and evaluate the structure of scientific collaboration.
• Identify influential authors and core research clusters.

---

🧩 Dataset

(Dữ liệu)

The dataset was collected through OpenAlex API, focusing on the Data Science domain (concept ID: C2522767166).
(Dữ liệu thu thập qua OpenAlex API, tập trung vào lĩnh vực Khoa học Dữ liệu.)

Key fields included:

• Work_ID
• Title
• Year
• Cited_by (citation count)
• Authors, Author_IDs
• Concepts

Example cleaned dataset (from paper):

Work_ID	Title	Year	Cited_by	Authors	Author_IDs
W2144634347	Molecular Cloning: A Laboratory Manual	2001	133,517	Joseph Sambrook; Elisabeth Fritsch; Tom Maniatis	A5112152140; ...
W4300870773	Statistical Power Analysis for the Behavioral Sciences	1989	83,956	Keith E. Muller; Jacob Cohen	A5110163574; ...

(Bảng minh họa dữ liệu đã được làm sạch.)

---

🏗 Methodology

(Phương pháp nghiên cứu)

1. Data Collection & Preprocessing

• Collect Data Science papers via OpenAlex API.
• Remove duplicates & missing values.
• Extract author list → build co-author edges.
• Normalize author identifiers.

2. Build Co-Author Network

• Graph G = (V, E) where:
  • Node = Author
  • Edge = Co-authorship
  • Weight = # of collaborations

3. Apply Community Detection Algorithms

• Louvain
• Leiden
• Fast Greedy (Clauset–Newman–Moore)

4. Compute Centrality Scores

• Degree
• Betweenness
• Closeness
• Eigenvector

5. Visualization

• Colored communities
• Node size scaled by centrality
• Force-directed graph layout

---

📊 Co-Author Network Statistics

(Thống kê mạng đồng tác giả)

Metric	Value
Nodes	9,212
Edges	136,198
Density	0.0032
Avg Degree	29.57
Clustering Coefficient	0.8608

(Cho thấy đây là mạng lớn, thưa nhưng liên kết nội bộ mạnh.)

---

🥇 Community Detection Comparison

(So sánh thuật toán phát hiện cộng đồng)

📌 Comparison Table

(Bảng so sánh)

Algorithm	Communities	Modularity (Q)	Strengths	Weaknesses
Louvain	1,167	0.9416	Fast, scalable, high modularity	Resolution limit
Leiden	1,167	⭐ 0.9418 (highest)	Well-connected communities, stable	Slightly more complex
Fast Greedy	1,164	0.9344	Produces hierarchy (dendrogram)	Lower modularity, slower on big graphs

⭐ Summary

• Leiden achieves the best modularity and stability.
• Louvain is nearly equal, extremely fast.
• Fast Greedy is good for hierarchical analysis.

(Leiden tốt nhất, Louvain nhanh nhất, Fast Greedy tốt cho phân cấp.)

---

🧠 Centrality Analysis

(Phân tích chỉ số trung tâm)

Authors with highest:

• Degree Centrality: Michelle Giglio, Ricardo Silva, Pascale Gaudet
• Betweenness Centrality: John P. A. Ioannidis (top “bridge”)
• Closeness Centrality: Asif Chinyavalla, Sandra W. Clifton
• Eigenvector Centrality: Chris Mungall, Karen Christie

(Những tác giả có tầm ảnh hưởng nhất theo từng chỉ số.)

---

📈 Visualizations

(Trực quan hóa)

This project visualizes:

• Top 2000-author network
• Largest 10 communities
• Centrality distributions
• Algorithm comparison charts

(Biểu đồ cấu trúc mạng, cộng đồng và chỉ số ảnh hưởng.)

---

📚 Discussion

(Thảo luận)

Key findings:

• Data Science research forms hub–spoke structures.
• Strong community separation (Q ≈ 0.94).
• Some authors act as “knowledge bridges” across areas.
• Leiden is the most stable and accurate method.
• Fast Greedy is useful for hierarchical clustering.

(Các cộng đồng lớn hoạt động như hạt nhân lan tỏa tri thức.)

---

🏁 Conclusion

(Kết luận)

This research demonstrates the effectiveness of combining:

• community detection,
• centrality analysis,
• scientific collaboration modeling

to uncover hidden structures in the Data Science ecosystem.

Future extensions may include:

• dynamic network analysis,
• topic modeling integration,
• graph neural networks (GNNs).

---

📄 Reference

(Tài liệu tham khảo — xem đầy đủ trong PDF)
Source: full report extracted from uploaded file. :contentReference[oaicite:2]{index=2}

---