/Anomaly-Detection-Using-Autoencoders

Autoencoder Neural Network is trained on credit card transaction data to detect anomalous transactions in near real time using flask api

Primary LanguageJupyter Notebook

Anomaly Detection Using Autoencoders

Credit Card Fraud Dection

Overview

Business Problem

  1. Objective

    • To assess if a transaction is fraudulent from the given credit card data
    • Learn from various features of normal transactions to distinguish fraudulent transactions

  2. Machine Learning Problem

    • Develop a machine learning model based on deep auto-encoders to learn distribution and relation between the features of normal transactions

  3. Technology

    • Python, Scikit-learn, TensorFlow, Keras, Pandas, Numpy

  4. Metrcs

    • Metrics: MSE (Mean Squared Error)

  5. Deployment

    • Deploy model in a scalable way so that business decisions can be taken in near real time in assessing riskiness of a transaction

  6. Approach

    • Exploratory Data Analysis
    • Data Cleaning
    • Build a base auto-encoder model using Keras
    • Evaluate and Tune the model
    • Make Predictions

  7. Deployment

    • Serve model as API endpoint using Flask
    • Perform real-time predictions

python Engine.py

 Train - 0
 Predict - 1
 Deploy - 2

Enter your value: 0

Data loaded into pandas dataframe
Preprocessing has begun...
Data cleaning has completed...
Data normalization has completed...
Preprocesing is complete...