Online Payments Fraud Detection ML Model

"Online Payments Fraud Detection ML Model: Flask-framework based App". The training notebooks & the dataset-link, outputs and sample-video are also provided in the respective folders with deployment.

Comment

Inspiration

  • To learn about the flask, streamlit deployment frameworks to deploy the machine learning model.

What it does

  • It can predict whether the transaction is either Fraud or not.

How we built it

Requirements:

  • Python 3.9+
  • Jupyter Notebook installed.

What i have done

  • Load the dataset which contains 6362620 entries in it and having 11 features in it.
  • Performing EDA on the dataset to get insights of the dataset.
  • Plotting different features graphs correspond to target feature.
  • Analyse the dataset by using correlation and plot the bar plot i.e., how much it is related to target feature.
  • Reduce the parameters and split the dataset into input and target features.
  • Split the parameters into training and testing sets.
  • Train the different models and get their accuracies and MSE & R2 scores even after tuning the hyper-parameters.
  • Even build a neural network and tune the parameters of their.
  • But Decision Tree Classifier Model gives promising performance on this dataset and classify and fit to the target variable with upto 99.97%.
  • Save the model into .joblib extension file and create a front-end for it.
  • Also creating a requirements.txt file for the model and website build-up.
  • Create a front-end using FLASK framework and create a user-friendly template.
  • Website can takes input and pass to the backend of the model and model will predict and provide the user a best result as of accuracy is around 99.97%.

Challenges we ran into

  • To learn in the pathway of flask framework, main is how to integrate the web framework with model.
  • How to integrate the model which is built on kaggle and downloaded model with website.
  • .pickle file is not working as expected so using .joblib file.

Accomplishments that we're proud of

  • Build a integrate and responsive web framework for users.
  • Unit test performed which also verified.
  • Build a perfect readme file which helps users to easily accessible and understandable to non-code users also.

What we learned

  • Flask
  • ML model deployment
  • Integration web framework with ML Model
  • New pipelines of ml model
  • About .pickle and .joblib files.

What's next for Online Payments Fraud Detection ML Model

  • Model deployment takes money on many platforms.
  • So, thinking to deploy the model on some platform and create an accessible link for the users.
Share this project:

Updates