🍃Dimensional Reduction
When visualizing data, each column represents a dimension. For instance, data plotted on the x-axis and y-axis is considered 2-dimensional, while data plotted on the x, y, and z axes is 3-dimensional.
Plotting data on a Cartesian coordinate system (x, y) is a simple and intuitive way to understand it. However, visualizing data with more than 3 dimensions becomes challenging. Thanks to the power of the Crystal Ball, Lady H. can project higher-dimensional data into lower dimensions while preserving as much of the original information as possible. This ability is known as "Dimensional Reduction".
About the Data
The data input used here is our Garden Bank's campaign data. Before dimensional reduction, we need to do some data preprocessing first:
Categorical features need to be coverted into numerical values, so that dimensional reduction algorithms can consume them. A common practice is through "encoding", such as one-hot encoding. Lady H. often uses Target Encoding, a technique that can sometimes enhance model performance. This method converts categorical values into numerical ones by incorporating both the prior knowledge of the target across all training data and the posterior knowledge of the target given each specific categorical value. See the results of this encoding below:
The next step is to standardize all numerical features to the same scale, ensuring that features with larger ranges do not dominate the distance metrics. This step is crucial before applying most dimensional reduction algorithms, as they rely on Euclidean distances.
Last updated