-
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
-
Machine Learning Problem
- Develop a machine learning model based on deep auto-encoders to learn distribution and relation between the features of normal transactions
-
Technology
- Python, Scikit-learn, TensorFlow, Keras, Pandas, Numpy
-
Metrcs
- Metrics: MSE (Mean Squared Error)
-
Deployment
- Deploy model in a scalable way so that business decisions can be taken in near real time in assessing riskiness of a transaction
-
Approach
- Exploratory Data Analysis
- Data Cleaning
- Build a base auto-encoder model using Keras
- Evaluate and Tune the model
- Make Predictions
-
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...