Understanding the importance of data

Many algorithmic problems for predictions and model fitting are hard to model, compute, and optimize using classic optimization algorithms or complex heuristics. Supervised machine learning provides a powerful new way to solve the most complex problems using optimization and a ton of labeled training data. The more data there is, the better the model.

One important thing to remember when working with ML algorithms is that models are powered by the training data you provide them and the training labels. Good data is the key to good performance. By data, we usually mean training data and using label annotations, one of the most notorious but also most important tasks in an ML project.

In most ML projects, you'll spend over 75% of the time with data analysis, preprocessing, and feature engineering. Understanding your data inside and out is critical to developing a successful predictive model. Think about it this way—the only thing...