"# Predicting Credit Card Fraud\n\n\nThe goal for this analysis is to predict credit card fraud in the transactional data. Fraud data is very highly skewed data that is, overwhelming majority of the data is one class and the rest is the other class. If we try to predict the majority class, the overall accuracy may be high but we may completely miss the other class that is, the fraud transactions here. \nI am using tensorflow to build the predictive model, and then t-SNE to visualize the dataset in two dimensions. See references for the detail help taken in this analysis.\n\nThe sections of this analysis include:\n\n* Exploring the Data \n* Building the Neural Network \n * FC network with Tensorflow\n * FC network with Keras\n * Autoencoder network with Keras: ** detects Fraud transaction**\n* Visualizing the Data with t-SNE."

FilesExpand file tree

tfnn-reference.ipynb

Latest commit

History

tfnn-reference.ipynb

File metadata and controls