Authors
Palmov S.
|
Degree
|
PhD in Technique, Associate Professor, State University of Telecommunications and Informatics, Samara State Technical University |
|---|---|
|
E-mail
|
psv@psuti.ru |
|
Location
|
Samara |
|
Articles
|
Data Mining applicability study for the telecommunications company’s customers analysisThe competitiveness of a company is influenced by many factors. A person, very often, cannot effectively
take them into account and make a balanced solution. This fact negatively impacts on business.
One way to correct the aforementioned situation is the usage of artificial intelligence, in particular,
Data Mining. Russian companies, unfortunately, rarely use the technology. The purpose of the
paper was to demonstrate the capabilities of Data Mining in terms of improving the company’s competitiveness.
To the effect, a file with depersonalized client information of a real telecommunications
company was used. In addition, the authors analyzed the prognostic abilities of Data Mining methods
to identify the one that best suits for the specified subject area. In this paper, using the Orange analytical
system, six methods were tested: «Decision Tree», kNN, «Random Forest», SVM, «Neural Network»,
and «Association Rules». Each of them was tested in two stages. At the first stage, with the fixed values
of the «Evaluate» widget settings (see Table 3), the prognostic model of the selected Data Mining
method was studied. At the second stage, the values of the «Evaluate» widget settings (see Tables 4
and 5) were changed, and the values of the predictive model settings were fixed (most effective values
of the settings from the previous step was used). An F-measure was used to evaluate the model’s performance.
As a result, it was found that the «Random Forest» and SVM are the most preferred methods.
Read more...
|
