silhouette
Silhouette cluster analysis implementation in Go
What It Does
Silhouette refers to an algorithm used to interpret and validate the consistency within clusters of data.
The silhouette value is a measure of how similar an object is to its own cluster compared to other clusters. The silhouette ranges from −1 to +1, where a high value indicates that the object is well matched to its own cluster and poorly matched to neighboring clusters.
If most objects have a high value, then the clustering configuration is appropriate. If many points have a low or negative value, then the clustering configuration may have too many or too few clusters.
When You Should Use It
- When you have numeric, multi-dimensional data sets
- If you want to check whether your data set is clustered
- When you have a vague idea of the clustering in your data set
- You want to figure out the optimal clustering configuration
Example
import (
"github.com/muesli/silhouette"
"github.com/muesli/clusters"
"github.com/muesli/kmeans"
)
// initialize your data set
// for the example we'll use three distinct clusters of data points
var d clusters.Observations
for x := 0; x < 64; x++ {
d = append(d, clusters.Coordinates{
rand.Float64() * 0.1,
rand.Float64() * 0.1,
})
}
for x := 0; x < 64; x++ {
d = append(d, clusters.Coordinates{
0.5 + rand.Float64()*0.1,
0.5 + rand.Float64()*0.1,
})
}
for x := 0; x < 64; x++ {
d = append(d, clusters.Coordinates{
0.9 + rand.Float64()*0.1,
0.9 + rand.Float64()*0.1,
})
}
// silhouette will theoretically work with multiple clustering algorithms
// it's commonly used with k-means
km := kmeans.New()
// compute the average silhouette score (coefficient) for 2 to 8 clusters, using
// the k-means clustering algorithm
scores, err := silhouette.Scores(d, 8, km)
for _, s := range scores {
fmt.Printf("k: %d (score: %.2f)\n", s.K, s.Score)
}
// estimate the amount of clusters in our data set
// this returns the k with the highest score (where 2 <= k <= 8)
k, score, err := silhouette.EstimateK(d, 8, km)
// k is usually 3 for this example, with a score close to 1.0
// note that k-means doesn't always converge optimally
...
}