Cluster analysis of internal migration processes between federal districts of the Russian Federation

Abstract

In this article, the author investigated migration processes between the federal districts of the Russian Federation for the period from 2012 to 2016 using the method of cluster analysis and presented the result of the analysis. An open data source was used for the study - the website of the Federal State Statistics Service. Socio-economic indicators of the federal districts of the Russian Federation were selected. A set of selected factors was investigated using correlation analysis to remove multicolliner factors. The processed set was then used for cluster analysis. The class analysis was carried out using machine learning methods using the Python programming language in the Jupyter Notebook development environment. The KMeans algorithm (k-means method) was used for the analysis. To understand how many clusters should be indicated, hierarchical clustering was carried out using the Ward method, in which the distances between clusters are an increase in the sum of the squares of the distances of objects to the cluster centers obtained as a result of their union.It was found that the original set is divided into three clusters. These results were used in the work of the KMeans algorithm, which divided the entire data set into three clusters and assigned a label to each row of data. Next, the average value of each socio-economic indicator in each cluster was calculated. Then, using the RandomForestClassifier classification algorithm, the significance of each factor was evaluated. As a result of cluster analysis, a set of features was obtained - socio-economic indicators of the region that affect the inclusion of the district in one or another migration class. Thus, we have obtained a set of controlling factors that can help to adjust migration flows in the Russian Federation. It was also found that there is a third type of regions, in addition to receiving and "donor" regions.