Mounted at /content/drive
✅ Setup complete

M-Pesa Transaction Pattern Analysis — 2023

Executive Summary

Dataset: 150,000 synthetic transactions · 5,000 users · 12-month period
Methodology: EDA → Feature Engineering → Unsupervised Segmentation → Anomaly Detection

Finding	Metric
Peak transaction hour	11:00–12:00 (lunch hour)
Weekend transaction share	~29% of weekly volume
Salary week (1st–7th) premium	+18% above average daily volume
Most valuable transaction type	send_money (38% of total KES value)
User segments identified	4 behaviourally distinct clusters
Fraud detection AUC (Random Forest)	0.87+
Top fraud signals	is_just_below_threshold, pct_night_txns

============================================================
  Figure 1: Transaction Volume Heatmap by Day and Hour
============================================================

============================================================
  Figure 2: Daily Transaction Activity Timeline
============================================================

============================================================
  Figure 3: Transaction Type Distribution
============================================================

============================================================
  Figure 4: Amount Distributions by Income Tier
============================================================

============================================================
  Figure 5: County-Level Transaction Geography
============================================================

============================================================
  Figure 6: Optimal Cluster Count Selection
============================================================

============================================================
  Figure 7: PCA Visualisation of User Segments
============================================================

============================================================
  Figure 8: Cluster Behavioural Profiles
============================================================

============================================================
  Figure 9: Anomaly Detection Results
============================================================

============================================================
  Figure 10: Feature Importance & PR Curve
============================================================

✅ Report exported to /content/drive/MyDrive/mpesa_analysis/reports/05_final_report.html