9. Cluster Analysis
K-means++
•Randomly select a point from the set of data
points entered as the first cluster centroid.
•For each set of data points x, calculate its
distance to the nearest cluster centroids
•Select a new data point as the new cluster
centroid, the principle choice is the more large
distance point
•Repeat until k cluster centroids have been
chosen
•Now that the initial centroids have been
chosen, proceed using standard k-means
clustering.
10. Cluster Analysis
•Assign each sample in each cluster.
•Compute distance between two clusters
•Combine the two nearest clusters
•Repeat, until all samples are assigning into
one cluster.
Agglomerative
Clustering
15. Prediction
Markov
Chains
•Network KPI state space
S = {99%~100%, 95%~99%, 90%~95%,
85%~90%, 80%~85%, <80%}
S = {1,2,3,4,5,6}
•Transition probability matrix (T)
•Initial state probability vector (P0)
19. Conclusion
Merged two methods for feature selection in
unsupervised learning
Four clustering models
Predicting by Markov chains
Evaluating clustering models