This chapter covers
- A comparison of machine learning and deep learning as methods to tackle tabular data problems
- A comparison of machine learning and deep learning in terms of simplicity
- A comparison of machine learning and deep learning in terms of transparency
- A comparison of machine learning and deep learning in terms of efficacy
There’s an open debate in the data science community about the best machine learning approach for tabular data. Some assert that classic machine learning techniques, such as gradient boosting using tools like XGBoost or LightGBM, are superior for most tabular data problems. Others advocate for including deep learning in your analysis toolkit. In this chapter, we’ll examine these two approaches using two concrete examples:
- Predict the price of Airbnb New York City listings. In this example, we use a real-world dataset of Airbnb listings to train models that predict whether a new listing will have a price above or below the average listing price. We’ll use this example to examine simplicity, transparency, and efficacy.
- Predict the length of time a property in a local real estate market is on the market before it is sold. In this example, we use a contrived real estate listing dataset to illustrate the explainability aspect of transparency.
We will focus on three criteria that are of particular value in interpreting scientific and business data: