CST 3170 Home Page The Prior Page The Next Page
Middlesex Logo

Lab 5: Case-Based Reasoning

  1. Make a CBR system for recommending plants for a garden.
  2. As ever, we will do this in eclipse in Java. Open up a new project. You should probably make it just text mode.
  3. Define several cases. Let's start by defining the garden by two features, height and width. The output is a third feature a plant.
  4. For example 5x5 -> holly; 10x10->maple; 20x20-> oak.
  5. Now make a nearness metric using standard euclidean distance.
  6. Test on a hard coded query (e.g. 4x4).
  7. Now read in the query from the user.
  8. Now add a third input feature, largest existing shrub or tree.
  9. Modify your similarity metric to include this.
  10. Increase the size of your casebase. (It would be good to have 20 cases.)
  11. Can you add a fourth feature (e.g. number of existing large shrubs)?
  12. Can you make the similarity metric hierarchical (e.g. maple and oaks are large trees, holly a small tree, and rose a shrub.)?