inf-428-data-analytics-online

K Nearest Neighbors

References

Wikipedia
Youtube Scikitlearn

K Nearest Neighbors

• Start with a set of example points
• The goal is to classify new points
  • For every example point
• Calculate Geometric Distance to new point
• Sort distances in decreasing order
• Take k smallest distances
• Use majority class as prediction for new point

K Nearest Neighbor in KNIME

On the KNIME Server there is an example in the ‘Machine_Learning’ folder. Get the example from the server and open it. Your KNIME should look as follows.

The video here shows how to build the workflow.

• File Reader
  • Read the data
• Color Manager and Scatter Plot
  • Look for patterns and relationships
• Partitioning
• Divide into Training and Testing
• Normalizer (Optional)
• Precondition data so every feature has similar weight
• K Nearest Neighbor
• Inputs are training and testing data
• Scorer
• Test how the learner performed

K Nearest Neighbor in Python

This notebook shows how to run K Nearest Neighbors in Python.

• Uses scikit learn library
• Steps are the same as KNIME
  • Divide into Training and Testing
  • Train a classifier
  • Apply classifier to test set
  • Score