Project 3 - Classification Using Non-Parametric Density Estimation and Performance
Evaluation -
Basic requirement (80)
Task 1 (20 pts): Implement kNN
The implementation should be able to flexibly change the value of "k"
The implementation should be able to measure the run time
Task 2 (20 pts): Experimenting kNN using the pima dataset
Apply kNN on nX, pX, and fX (refer to project 2) and compare the performance with that from
project 2 from both computational time and accuracy perspectives
For each appliation, try different values of k, e.g., k=1, 5, and sqrt(n). Use Euclidean and other
Minkowski distance of different degrees (Refer to Lecture note 10).
Task 3 (20 pts): Use the "fglass" data set. Experimenting cross-validation with kNN as the classifier. Try
out different k's. Based on the performance from cross-validation, determine the best k for this data set.
You need to get enough samples of k between 1 and sqrt(n), e.g., 1, 5...15, and sqrt(n).
Task 4 (20 pts): Implement NB classifier fusion approach and report the best fusion result.
Report (20)
Bonus (+15)
Design and implement methods other than discussed in class to reduce the computational load of
kNN. Show the time spent and the storage used in the new implementation. Also compare the
accuracy.
