2.4 Feature Engineering

2.4 Feature Engineering

Li Mu

Station B: https://space.bilibili.com/1567748478/channel/collectiondetail?sid=28144
Course homepage: https://c.d2l.ai/stanford-cs329p/

1. Why feature engineering is needed:

• feature engineering

  Feature extraction is performed on the data set, so that the machine learning model can learn faster, with higher accuracy and better effect when learning the data processed by feature engineering.

• Why Feature Engineering?

  Before deep learning became popular, traditional machine learning models were commonly used. Therefore, before learning, the data needs to be processed into the data form that the model likes (this process is often designed manually by people), because machine learning algorithms are more "likes" fixed-length input and output. This is a very critical technology. For example, in computer vision, images are often processed into a vector to train an SVM model. After the deep learning technology matured, people began to use neural networks for feature extraction ( making feature engineering easier ), but this process and status of feature engineering did not change . Moreover, the neural network can continuously change parameters to better extract features, but the disadvantage is that it requires a lot of data and resources.

2. Tabular data features:

• in