Lab 20: Nearest Neighbour Clustering
- Grab the data set from
which is the S1 of S-sets example.
- Write a program (C# works) to cluster these items.
- Read the data into an array. (2 points)
- Set K points randomly (you can try K=15 or 20). (2 points)
- Assign each point to one of the K points. (1 point)
- Move K to the centre of its points. (1 point)
- Repeat until none change. (2 points)
- Print result. (1 point)
- Compare (manually using excel ok) result. (1 point)
- If you read the datasets file, you'll notice that
there are 15 clusters.
- You'll want to start by reading in the data.
- I used kMeans with K = 15.
- On my first test, I got 3372/5000. (Some clusters weren't really
- I used excel to measure (the answers are in datasets above).
- Can you measure?
- I did have problems with the distance and squares overflowing
the fields. So, I had to decompose the distance equation into
parts and combine them with floats. (That took longer than the
rest of the problem.)