BU MET CS-677: Data Science With Python, v.2.0 CS-677 Assignment: k-means
Assignment
Implement k-means clustering. Your objects are weeks and
your feature set is (µ, σ) for your weeks. We will choose the
optimal number of clusters using year 1 and year 2 data. We
apply the classifier to our data and will examine the composi-
tion of your clusters
Questions:
1. take k = 3 and use k-means sklearn library routing for k-
means (random initialization and use the defaults). Take
k = 1, 2, . . . 7, 8 and compute the distortion vs. k. Use the
”knee” method to find out the best k.
2. for this optimal k, examine your clusters and for each clus-
ter compute the percentage of ”green” and ”red” weeks in
that cluster.
3. does your k-means clustering find any ”pure” clusters (per-
cent of red or green weeks in a cluster is more than, say,
90% of all weeks in that cluster)?
Page 1