We created a customer churn model in Chapter 3, Building Deep Neural Networks for Binary Classification, that is capable of predicting whether a customer will leave an organization based on specified data. We might want to train the existing model on newly available data. Transfer learning occurs when an existing model is exposed to fresh training on a similar model. We used the ModelSerializer class to save the model after training the neural network. We used a feed-forward network architecture to build a customer retention model. 

In this recipe, we will import an existing customer retention model and further optimize it using the DL4J transfer learning API.