There are 11,500 rows in this dataset. Each row of the dataset contains 178 data points, each representing a 1-second sample of an EEG recording and a corresponding patient state, generated across 100 different patients. 

There are five patient states in the dataset; however, patients in state 2 through state 5 were not experiencing a seizure. Patients in state 1 were experiencing a seizure.   

I have modified the original dataset, reframing the problem into a binary classification problem by changing states 2-5 to class 0, which will mean no seizure and class 1, which will mean seizure.

As with the regression problem in Chapter 2, Using Deep Learning to Solve Regression Problems, we will be using an 80% train, 10% val, 10% test split.