Features handling

Note

You can change the settings for feature processing under Models > Settings > Features tab

Most machine learning engines in DSS visual machine learning can only process numerical features, with no missing values.

DSS allows users to specify pre-processing of variables before model training.

• Features roles and types
  • Roles
  • Variable type
• Categorical variables
  • Category handling
    • Target encoding
      • Impact coding
      • GLMM encoding
    • Ordinal encoding
    • Frequency encoding
  • Missing values
• Numerical variables
  • Numerical handling
    • Datetime cyclical encoding
  • Rescaling
  • Missing values
• Text variables
  • Text handling
    • Text embedding
    • Using models from the code environment resources
    • Using models from the LLM Mesh
  • Missing values
• Vector variables
  • Missing values
• Image variables
  • Image handling
    • Image embedding
  • Missing values
• Custom Preprocessing
  • Implementing a custom processor
    • Example
  • Naming output columns
    • Example