COMP7108 Project: Graph Databases for Financial Intelligence

This repository contains our COMP7108 project implementation using the PaySim transaction dataset and Neo4j/Cypher for fraud-oriented graph analytics.

1) Project Scope

The project covers the assignment pipeline end-to-end:

• Property-graph modeling and ingestion of PaySim data
• Fraud pattern discovery with graph traversal queries
• Fraud-subgraph visualization
• Graph algorithm analysis (GDS)
• Temporal fraud behavior analysis

2) Dataset

• PaySim (Kaggle): https://www.kaggle.com/datasets/ealaxi/paysim1
• Source note: scripts/dataset-source.txt

Note: the raw CSV is not included in this repository. Please download it and place it in Neo4j's import-accessible location.

3) Repository Structure

• Project/taskA: ingestion, schema reset, constraints/indexes, validation
• Project/taskB: pattern queries (fan-in, fan-out, mule, cycle attempts, shortest path)
• Project/taskC: fraud subgraph visualization query + exported figures
• Project/taskD: graph algorithm queries (PageRank, community, betweenness)
• Project/taskE: temporal-analysis outputs and Python post-processing scripts
• scripts: assignment brief and project guidance documents

4) Environment Requirements

• Neo4j (for Cypher execution)
• Neo4j Graph Data Science plugin (required for Task D queries)
• Python 3.9+ (for Task E scripts)
• Python packages:
  • pandas
  • matplotlib

5) How to Run

Step A. Prepare data for import

1. Download PaySim CSV from Kaggle.
2. Rename/place the file so that Cypher import path in Project/taskA/taskA03_import.cypher is valid (file:///data.csv by default).

Step B. Run Task A (in Neo4j Browser, in order)

1. Project/taskA/taskA01_reset_and_schema.cypher
2. Project/taskA/taskA02_constraints_and_indexes.cypher
3. Project/taskA/taskA03_import.cypher
4. Project/taskA/taskA04_validation_queries.cypher

Step C. Run Task B/C/D queries

• Task B queries:
  • Project/taskB/taskB01_fan_in.cypher
  • Project/taskB/taskB02_fan_out.cypher
  • Project/taskB/taskB03_mule_detection.cypher
  • Project/taskB/taskB04_cycle_detection.cypher
  • Project/taskB/taskB05_shortest_path_and_exploration.cypher
• Task C visualization:
  • Project/taskC/taskC_visual.cypher
• Task D algorithms (GDS required):
  • Project/taskD/taskD01_pagerank.cypher
  • Project/taskD/taskD02_Community_Betweenness.cypher

Step D. Run Task E analysis

1. (E01) Run Project/taskE/taskE01_transaction_spike.cypher and export the results as taskE01_outgoing.csv and taskE01_incoming.csv.
2. (E01) Edit paths in Project/taskE/taskE01_transaction_spike.py to your local files, then run the script to generate spike results.
3. (E02) Run Project/taskE/taskE02_balance_anomaly.cypher and export anomaly records.
4. (E03) Run Project/taskE/taskE03_fraud_timeline.cypher and export timeline CSVs.
5. (E03) For fraud timeline plotting, use Project/taskE/taskE03_fraud_timeline_plot.py with Project/taskE/output/taskE03_fraud_timeline.csv.

Important: current Python scripts contain Windows absolute paths in the source code; update them before execution.

6) Mapping to Assignment Requirements (Task A–E)

Task A: Data ingestion and optimization

• Reset/schema initialization: Project/taskA/taskA01_reset_and_schema.cypher
• Constraints and indexes (optimized lookup/filtering): Project/taskA/taskA02_constraints_and_indexes.cypher
• Batch import (IN TRANSACTIONS OF 1000 ROWS): Project/taskA/taskA03_import.cypher
• Validation checks (counts, distributions, fraud stats): Project/taskA/taskA04_validation_queries.cypher

Task B: Pattern discovery

• Fan-in detection: Project/taskB/taskB01_fan_in.cypher
• Fan-out detection: Project/taskB/taskB02_fan_out.cypher
• Mule pattern detection: Project/taskB/taskB03_mule_detection.cypher
• Cycle detection attempts: Project/taskB/taskB04_cycle_detection.cypher
• Shortest path and local path exploration: Project/taskB/taskB05_shortest_path_and_exploration.cypher

Task C: Visualization

• Fraud account labeling and subgraph extraction: Project/taskC/taskC_visual.cypher
• Exported figures: Project/taskC/C_step.png, Project/taskC/C_amount.png

Task D: Graph algorithm application

• PageRank ranking and fraud cross-check: Project/taskD/taskD01_pagerank.cypher
• Community detection + betweenness on local suspicious structure: Project/taskD/taskD02_Community_Betweenness.cypher
• Supporting result tables: CSV files in Project/taskD

Task E: Temporal analysis

• E01 — Transaction spike detection:
  • Cypher export query: Project/taskE/taskE01_transaction_spike.cypher
  • Python spike analysis script: Project/taskE/taskE01_transaction_spike.py
  • Results: Project/taskE/output/taskE01_outgoing.csv, Project/taskE/output/taskE01_incoming.csv
  • Top spike accounts: Project/taskE/output/taskE01_outgoing_spikes.csv, Project/taskE/output/taskE01_incoming_spikes.csv
• E02 — Balance anomaly detection:
  • Cypher query: Project/taskE/taskE02_balance_anomaly.cypher
  • Detailed anomaly records: Project/taskE/output/taskE02_detailed_anomaly_records.csv
  • Summary by transaction type: Project/taskE/output/taskE02_summary_transaction_type.csv
• E03 — Fraud activity timeline:
  • Cypher queries (fraud counts per step + account activity span): Project/taskE/taskE03_fraud_timeline.cypher
  • Python plot script: Project/taskE/taskE03_fraud_timeline_plot.py
  • Timeline data: Project/taskE/output/taskE03_fraud_timeline.csv, Project/taskE/output/taskE03_flagged_timeline.csv
  • Account activity span: Project/taskE/output/taskE03_account_activity.csv
  • Timeline chart: Project/taskE/output/taskE03_fraud_timeline.png

7) Deliverables in This Repository

Aligned with assignment deliverables:

• Ingestion scripts: Project/taskA
• Query collection: Project/taskB, Project/taskC, Project/taskD, Project/taskE
• Visualization outputs: Project/taskB/output, Project/taskC, Project/taskE
• Algorithm/temporal outputs: CSV/PNG files under Project/taskD and Project/taskE