My main goal in this project was to see how well Keras can be used to classify binary. First, I wanted to perform a visual analysis to discover if there are any trends that can be spotted from a human perspective and point out features of interest. Finally, after labeling and feeding the data to the model, I assessed the performance of this approach.
For the development of this project I retrieved data from - Kaggle.
- Visualizing data via Matplotlib and Plotly to discover any trends that would be present through breaking down data visually. Horizontal Bar Chart matrix was used to represent distinct counts of each feature per category and Parallel Category Chart was used to see the feature flow resulting in fraud/not fraud.
- Encoding columns via cat.codes and splitting data for the model fitting.
- Creating a Keras model with 32 input neurons in the input layer and 32 neurons in 1 hidden layer, and utilizing adam optimizer and binary crossentropy.
- Evaluating model performance and achieving loss: 0.9092 - accuracy: 0.9361