In BuddyMove Dataset Data Set was populated from destination reviews published by 249 reviewers of holidayiq.com till October 2014. Reviews falling in 6 categories among destinations across South India were considered and the count of reviews in each category for every reviewer (traveler) is captured.
We dont use Attribute 1 : Unique user id and we cluster based on other features
- Attribute 1 : Unique user id
- Attribute 2 : Number of reviews on stadiums, sports complex, etc.
- Attribute 3 : Number of reviews on religious institutions
- Attribute 4 : Number of reviews on beach, lake, river, etc.
- Attribute 5 : Number of reviews on theatres, exhibitions, etc.
- Attribute 6 : Number of reviews on malls, shopping places, etc.
- Attribute 7 : Number of reviews on parks, picnic spots, etc.
Silhouette refers to a method of interpretation and validation of consistency within clusters of data. The technique provides a succinct graphical representation of how well each object lies within its cluster.
The silhouette value is a measure of how similar an object is to its own cluster (cohesion) compared to other clusters (separation). 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.
The silhouette can be calculated with any distance metric, such as the Euclidean distance or the Manhattan distance.
In this way we want to divide the data into k cluster and k as the input. Because we use classification so k equal to the number of classes in the classification.) k=6) Given the silhouette value obtained (0.2), we can conclude that the clustering is almost done well.
n_clusters : 6
silhouette_score : 0.2978550526393102
In this method, we want to divide the data into k-class, but in this way we do not have the number of classes as inputs. In this way, we use different linkage for clustring, and in each linkage, clustring is done for different k (1> k >10), and the best ones are selected. ) compared based on the silhouette.)
n_clusters : 2
silhouette_score : 0.24418093101151334
n_clusters : 3
silhouette_score : 0.28973852605254324
n_clusters : 4
silhouette_score : 0.24988027361895215
n_clusters : 5
silhouette_score : 0.24548845594378832
n_clusters : 6
silhouette_score : 0.2537695943060857
n_clusters : 7
silhouette_score : 0.26526788561620696
n_clusters : 8
silhouette_score : 0.2894292674564693
n_clusters : 9
silhouette_score : 0.2836963765775202
K=3 > k=8 > k=9 > k=7 > k=6 > k=4 > k=5 > k=2
In this case, the silhouette for k=3 has the highest value, so we can say that the k=3 is better.
n_clusters : 2
silhouette_score : 0.4763807066815255
n_clusters : 3
silhouette_score : 0.28849394043596843
n_clusters : 4
silhouette_score : 0.3247230508277392
n_clusters : 5
silhouette_score : 0.30359476855496775
n_clusters : 6
silhouette_score : 0.21029769932379913
n_clusters : 7
silhouette_score : 0.2569134280073912
n_clusters : 8
silhouette_score : 0.24288145683859752
n_clusters : 9
silhouette_score : 0.2451663065476037
K=2 > k=4 > k=5 > k=3 > k=7 > k=9 > k=8 > k=6
Based on the values, we can say that k=2 (s=0.4763) is better, because the silhouette has the highest value.
n_clusters : 2
silhouette_score : 0.4763807066815255
n_clusters : 3
silhouette_score : 0.24000100804992566
n_clusters : 4
silhouette_score : -0.028251777214029824
n_clusters : 5
silhouette_score : -0.10104153455087905
n_clusters : 6
silhouette_score : -0.11106803869427169
n_clusters : 7
silhouette_score : -0.2030951913529466
n_clusters : 8
silhouette_score : -0.08213424961087658
n_clusters : 9
silhouette_score : -0.07556247582712967
K=2 > k=3 > k=4 > k=9 > k=8 > k=5 > k=6 > k=7
Based on the values, we can say that k=2 (s=0.4763) is better. On the other hand, with regard to the values of the silhouette that are negated, we can say that they are not very good clustring.
n_clusters : 2
silhouette_score : 0.4763807066815255
n_clusters : 3
silhouette_score : 0.28849394043596843
n_clusters : 4
silhouette_score : 0.24955186842539795
n_clusters : 5
silhouette_score : 0.23637582388852013
n_clusters : 6
silhouette_score : 0.24591972732074335
n_clusters : 7
silhouette_score : 0.24807474873485053
n_clusters : 8
silhouette_score : 0.23478197924553293
n_clusters : 9
silhouette_score : 0.26757421494826006
K=2 > k=3 > k=9 > k=4 > k=7 > k=6 > k=5 > k=8
Based on the silhouette value obtained for number of different classes, we can conclude that k=2 (s=0.4763) that has the highest amount of silhouette is a better clustring. On the other hand, since in all cases the value of the silhouette is positive, we can say that all clustring are pretty good.
According to the silhouette in two ways we can say that hierarchical method is better because it has largest value in the silhouette score.