kagikush/Nigerian_telecom_fraud

---
language:
- en
tags:
- telecom
- fraud
- nigeria
- cyber
size_categories:
- 10K<n<100K
---
<video controls style="width: 100%; max-width: 720px;">
<source src="https://huggingface.co/datasets/kagikush/Nigerian_telecom_fraud/resolve/main/nigerian%20fraud%20-%20video%20exp.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>

# 🇳🇬 Nigerian Telecom Fraud Analysis

## Overview

This project presents an end-to-end Exploratory Data Analysis (EDA) of telecom-related fraud incidents in Nigeria. The goal is to uncover patterns, quantify financial impact, and provide actionable recommendations to mitigate fraud.

The analysis is based on a dataset containing ~50,000 fraud records, including fraud type, timestamps, operators, and financial impact.

---

## Objectives

* Understand the distribution of fraud types
* Identify the most financially damaging fraud categories
* Compare fraud activity across telecom operators
* Provide business-oriented recommendations

---

## Dataset Description

Key columns in the dataset:

* `fraud_id` – Unique identifier for each fraud case
* `detected_at` – Timestamp when fraud was detected
* `fraud_type` – Category of fraud (e.g., phishing, account takeover)
* `operator` – Telecom operator involved
* `affected_phone` – Victim phone number
* `customer_id` – Customer identifier
* `financial_loss_ngn` – Estimated financial loss in Nigerian Naira

---

## Data Cleaning

Steps performed:

* Converted `detected_at` to datetime format
* Handled missing values in timestamps
* Standardized categorical values (fraud types, operators)
* Removed inconsistencies and duplicates where necessary

---

## Exploratory Data Analysis

### 1. Fraud Type Distribution

* The most common fraud types include:

* call forwarding fraud
* identity theft
* other frauds occur in Relatively similar frequency.


### 2. Financial Impact Analysis

* Highest average losses:

* **Account Takeover**
* **Subscription Fraud**
* Moderate impact:

* Phishing
* Call forwarding fraud
* Low impact:

* IMEI spoofing and similar technical frauds

Insight: Frequency does not necessarily equal severity. Some rare frauds are far more damaging.

### 3. Operator Comparison

* Fraud cases are distributed across multiple operators
* Some operators(MTN and Airtel) show higher exposure to specific fraud types

Insight: Targeted prevention strategies may be needed per operator.


---
### 3. Fraud location Comparison

* Fraud cases are distributed across multiple locations
* It is evident that most fraud cases occur in Lagos/

Insight: Lagos is seriously suspected of fraud.


## Key Insights

* A small number of fraud types account for the majority of financial loss
* High-frequency fraud is not always high-impact
* Fraud patterns suggest both opportunistic and organized activity

---

## Recommendations

### Focus on High-Impact Fraud

* Prioritize detection systems for:

* Account takeover
* Subscription fraud
* Allocate more resources to prevent high-loss incidents

### Improve Detection Systems

* Implement anomaly detection for unusual account behavior
* Use machine learning models for early fraud identification

### Strengthen Customer Protection

* Introduce multi-factor authentication (MFA)
* Send real-time alerts for suspicious activities

### Operator-Specific Strategies

* Customize fraud prevention measures per telecom operator
* Monitor operator-specific fraud trends

### Continuous Monitoring

* Build dashboards for real-time fraud tracking
* Regularly update models and rules

---

## Tools & Technologies

* Python (Pandas, NumPy)
* Data Visualization (Matplotlib, Seaborn)
* Jupyter Notebook

---

## How to Use

1. Clone the repository
2. Open the notebook in Jupyter
3. Run all cells sequentially
4. Explore visualizations and insights

---

## Future Work

* Build predictive models for fraud detection
* Incorporate real-time streaming data
* Add geolocation-based analysis
* Deploy as an interactive dashboard

---

## Contributing

Contributions are welcome! Feel free to open issues or submit pull requests.

---

## License

This project is for educational and analytical purposes.

---
## Data Source

The dataset used in this project was sourced from [nigerian-telecom-fraudulent-activity-datasets / huggingface].

## Acknowledgments

This analysis was conducted as part of a Data Science assignment at Reichman University. The dataset is used for educational purposes only.