Dimensionality reduction is a common step for data processing. This process is useful for feature engineering or data visualization. Too many features in a dataset can complicate data visualization and analysis afterwards. Therefore, dimensionality reduction is needed to overcome this problem.

Dimensionality reduction does not automatically reduce the existing features. This method will first summarize all features in a dataset into several components according to the algorithm we use. There are two different methods in dimensionality reduction: Linear (Principal Component Analysis) or Non-Linear (Manifold Learning). In this article we will implement and compare these two methods using well log data. Read more…