As part of the study of mobile device traffic, the formation of an experimental database of network traffic of selected mobile applications, which can be used for training and testing classifiers using machine learning methods, occupies an important place. To automate the process of studying mobile application traffic classification algorithms, a software package has been developed that allows you to automatically collect network traffic packets from mobile devices and save them to a database; group network traffic packets into streams; upon user request, generate data sets with specified characteristics (number of specific application flows: background traffic; generate a data set based on the already created set with the addition of new flows, excluding repetitions). To collect traffic from mobile devices running the Android operating system, an application was developed that uses the application programming interface to create virtual private networks, collects network traffic packets, identifies the source application and sends them via HTTP to server software. Using the client and server software, the database created was filled with traffic from 18 main applications. During the data collection, 71,667 streams and 6,989,991 packets were received.