Clustering includes one set - objects (X) and another - cluster numbers (Y). The distance function between objects (p) is given. The main goal is to split the training sample into clusters so that objects within each sector are close to each other by the metric p, and from different ones - differ significantly from each other. Each object is assigned a group number y(i). Clustering algorithm: a function that puts a cluster number Y in parallel with any object X.
What types of data are used in clustering?
The following are distinguished:
The described characteristics of objects are divided into numerical and non-numerical.
Table of distances between objects. Each targeted industry database object is described by its distances to all others in the clustering.
Only until 12/19
Download a selection of materials to be guaranteed to find a job in IT in 14 days
List of documents:
TOP-100 job search platforms from GeekBrains
20 professions of 2023, with income from 150,000 rubles
Checklist "How to Successfully Pass an Interview"
To receive the file, please enter your e-mail:
E-mail, for example, [email protected]
Please confirm that you are not a robot
by providing your phone number:
+7
912 345-67-89
Download the collection for freepdf 2.5mb
Already downloaded 52300
I confirm my consent to the processing of personal data .
How is clustering tested?
Evaluation of the results is a complex task, as is the clustering process itself. The most common methods include “internal” and “external” evaluation. In the first case, the system is reduced to a single qualitative indicator, and in the second, the clustering is compared with an existing classification or “ground truth”. Additionally, a human expert can conduct a manual evaluation and determine the usefulness of using the method in the intended application.
Clustering is a very useful tool, especially in the field of advertising data analysis. When it is necessary to effectively distribute the PR budget, attracting the maximum number of clients for the minimum cost, the method will help to determine the most appropriate approach.
