Getting ready

Given that the objective is to minimize error, let's define the error that we shall be minimizing—we should ensure that a positive error and a negative error do not cancel out each other. Hence, we shall minimize the absolute error. An alternative of this is to minimize the squared error.

Now that we have fine-tuned our objective, let's define our strategy of solving this problem:

• Normalize the input dataset so that all variables range between zero to one.
• Split the given data to train and test datasets.
• Initialize the hidden layer that connects the input of 13 variables to the output of one variable.
• Compile the model with the Adam optimizer, and define the loss function to minimize as the mean absolute error value.
• Fit the model.
• Make a prediction on the test dataset.
• Calculate the error in the prediction on the test dataset.

Now that we have defined our approach, let's go ahead and perform it in code in the next section.