2.1 Clustering
Cluster analysis is a classification technique for forming groups in a
complex set of data that does not rely on any presuppositions about the
number or structure of groups. In cluster analysis, membership in groups
is unknown for all observations, and even the number of groups is
unclear. The purpose of this technique is to identify homogeneous
groups. Groups are determined in such a way that the degree of
consistency between members of the same group is strong and the degree
of consistency among members of different groups is weak. Therefore,
cluster analysis is an effective tool in scientific and management
research that groups a set of data into a d-dimensional space so that
the similarity in the clusters is maximized and the consistency between
the two different clusters is minimized 9. Therefore,
clustering can be of great help to managers in selecting units for the
diagnostic network. In this study, we performed clustering based on the
efficiency score of the labs.
There are various clustering methods that are used in a wide range.
Among the clustering algorithms, the K-Means method is a very common
method for partitioning. The K-Means method is one of the methods of
data clustering in data mining. This method is a unique and flat method.
For this algorithm, different forms have been expressed. But all of them
have a repetitive process that attempts to obtain central points for
clusters in a number of fixed clusters, which are in fact the same as
the average of points belonging to each cluster, as well as to assign
each data sample to a cluster, in a way that the data sample has the
minimum distance to the center of that cluster. In the K-Means
algorithm, initially k members (k is the number of clusters) are
randomly selected from the n members as cluster centers. Then the n-k
remaining members are assigned to the nearest cluster. After assigning
all members, the cluster centers are recalculated and the members are
assigned to the clusters according to the new centers, and this
continues until the centers of each cluster remain constant. The best
clustering is one that maximizes the total similarity between the
cluster center and all cluster members, and minimizes the overall
similarity between cluster centers.
In this research, we use expert opinion to classify laboratory units
using the k-means method. It should be noted that the grouping is based
on the performance scores of the laboratory units.