Comparing LightGBM, XGBoost, and Deep Learning
The previous chapter introduced LightGBM for building gradient-boosted decision trees (GBDTs). In this chapter, we compare LightGBM against two other methods for modeling tabular data: XGBoost, another library for building gradient-boosted trees, and deep neural networks (DNNs), a state-of-the-art machine learning technique.
We compare LightGBM, XGBoost, and DNNs on two datasets, focusing on complexity, dataset preparation, model performance, and training time.
This chapter is aimed at advanced readers, and some understanding of deep learning is required. However, the primary purpose of the chapter is not to understand XGBoost or DNNs in detail (neither technique is used in subsequent chapters). Instead, by the end of the chapter, you should have some understanding of how competitive LightGBM is within the machine-learning landscape.
The main topics are as follows:
- An overview of XGBoost
- Deep learning and TabTransformers...
