Syllabus →

CT PYQ →

Pruning Strategies for Data Mining

Pruning in data mining refers to the process of reducing the size of a dataset or model by removing unnecessary or less significant parts. This helps improve efficiency, reduce complexity, and enhance the performance of data mining algorithms. Here are some common pruning strategies:

1. Pre-pruning (Early Stopping):

Description: Pre-pruning involves stopping the data mining algorithm before it becomes too complex, based on certain criteria. This is often used in decision tree algorithms.

Example:

• In decision tree construction, the tree growth can be halted if:
  • The number of instances in a node is below a threshold.
  • The depth of the tree exceeds a predefined limit.
  • Further splitting does not significantly improve the model accuracy.

Advantages:

• Reduces overfitting.
• Lowers computational cost.

Disadvantages:

• May lead to underfitting if the stopping criteria are too stringent.

2. Post-pruning:

Description: Post-pruning involves first allowing the algorithm to create a fully grown model, and then pruning back certain parts to reduce complexity.