/Fraud-Detection-in-Electricity-and-Gas-Consumption-in-tunisia

Primary LanguageJupyter NotebookMIT LicenseMIT

Fraud Detection in Electricity and Gas Consumption 🕵️

photo-by-american-public-power-association-on-unsplash

Project status: In Progress

Project Goal

To stop consumers from manipulating meters fraudulently and reduce financial losses incurred by the Tunisian Company of Electricity and Gas (STEG), assuring effective delivery of electricity and gas services across Tunisia.

Objective

To create a system that uses a customer's billing history to accurately identify and track down customers who are engaged in fraudulent activity. This will allow the Tunisian Company of Electricity and Gas (STEG) to take the necessary action to stop further losses and maintain the integrity of their services.

Data source:

The source of data for this project is the Fraud Detection in Electricity and Gas Consumption, which is available on Kaggle. The dataset consists of four distinct datasets, two of which are for testing and the other two for training. Both the training and testing datasets contain information on clients, as well as their billing history, covering a period from 2005 to 2019.

Technologies Used:

👉🏽 python - for data analysis and modeling

👉🏽 pandas - for data manipulation, visualization, and analysis

👉🏽 matplotlib -for data manipulation, visualization, and analysis

👉🏽 scikit-learn - for building predictive models

👉🏽 Numpy - for numerical computing in Python

👉🏽 Streamlit - For building web-based applications in Python for sharing models.

👉🏽 Seaborn -for data manipulation, visualization, and analysis

Models Used:

👉🏽 Logistic regression

👉🏽 Decision tree classifier

👉🏽 random forest classifier

👉🏽 gradient boosting classifier

👉🏽 K nearest neighbour

👉🏽 SGDClassifier

👉🏽 LGBMClassifier

👉🏽 AdaBoostRegressor

👉🏽 CatBoostClassifier

Best model:

Limitations:

  • Computational Complexity due to large number of data

Deployment

Follow these instructions to utilize the web app on your laptop:Clone the project repository to your local machine.

  • Install Python 3.7 (if not already installed).
  • Install the required dependencies by running pip install -r requirements.txt in your terminal/command prompt.
  • Navigate to the project directory and run streamlit run fraud.py to launch the web-based application. You should now be able to view the app on your local machine.😃 Here is the link to the deployed model