Build a Neural Network in Python (Binary Classification)

This is a cheat sheet for me to copy the code when I need it!

Photo by Peter Hansen on Unsplash

Set up the environment

Import modules that we are going to use

Set our data

We use Titanic: Machine Learning from Disaster as the example dataset

• Read the dataset
• Deal with the missing value
• Recode variables
• Check distribution of target variable

Prepare data for modeling

• Split the data(X,y)
• Convert to NumPy array

Build, Compile, Fit model

• Use the Sequential API to build your model
• Specify an optimizer (rmsprop or Adam)
• Set a loss function (binary_crossentropy)
• Fit the model (make a new variable called ‘history’ so you can evaluate the learning curves)
• EarlyStopping callbacks to prevent overfitting (patience of 10)

Evaluate the Model

• Learning curves (Loss)
• Learning curves (Accuracy)
• Confusion matrix