This is a short study of the heart rate and other characteristics of 2126 fetuses. The fetuses are classified in 10 classes of heart behavior, and 3 classes of disease risk ( NSP ). For more information on the parameters of the dataset, they are widely explained in the dataset description.
In a first part we will explore our dataset to try to find some valuable insights hidden in the numbers. To do that we will use dimensionality reduction, first through PCA, then using an autoencoder. We will analyze the results and compare them to our original dataset.
In a second part we will try to model a tool that could be useful in detecting the possible pathology. In order to do this, we will test a number of Machine Learning classifiers, we will try to find the most accurate for this problem and then tune it to its best accuracy. We will then compare the results of the best classifier we found with the results of a small artificial neural network and see which one does best.