To obtain access to full text of journal and articles you must register!
- Article name
- A machine learning model for solving regression problems related to identifying the actions of cheating bots
- Authors
- Pitelinsky K. V., , yekadath@gmail.com, Federal State Autonomous Educational Institution of Higher Education "Moscow Polytechnic University", Moscow, Russia
Britvina V. V., , saaturn2015@mail.ru, Moscow Polytechnic University, Moscow, Russia
Bobrova E. O., , kotei_katrin@mail.ru, Federal State Autonomous Educational Institution of Higher Education "Moscow Polytechnic University", Moscow, Russia
Kalutsky I. V., , kalutsky_igor@mail.ru, Federal State Autonomous Educational Institution of Higher Education "Moscow Polytechnic University", Moscow, Russia
Savenkov V. V., , savenckoa2015@gmail.com, Federal State Autonomous Educational Institution of Higher Education "Moscow Polytechnic University", Moscow, Russia
Agostinho A. C., , adcaculo@gmail.com, Moscow State University of Technology "STANKIN", Moscow, Russia
- Keywords
- cheater bots / website traffic / behavioral factors / information security / machine learning / neural networks / LSTM / attention layer / time series / forecasting / gradient boosting
- Year
- 2024 Issue 2 Pages 3 - 11
- Code EDN
- HHIUWS
- Code DOI
- 10.52190/2073-2600_2024_2_3
- Abstract
- A comprehensive study was carried out in the field of forecasting website traffic using machine learning. The importance of analyzing website traffic is noted for timely detection of the beginning of the malicious activity of bots that cheat behavioral factors, which ultimately reduce the rating of an information resource in search engines. A custom neural network (NN) architecture is proposed, based on LSTM layers and an attention layer, specially designed for processing time series. LSTM layers and the implementation of an attention layer to focus on key indicators are considered. The evolution of models has been studied, starting from simple recurrent neural networks and the introduction of the Embedding layer to the transition to more complex architectures. The stages of model training are presented, from selecting hyperparameters to evaluating results on test data in order to achieve a balance between performance and training resources. The model was assessed using the mean absolute and root mean square errors. The developed model is compared with gradient boosting and the advantages of NN are revealed. Prospects for future research and for improving the quality of analysis of website traffic dynamics using machine learning are indicated.
- Text
- BUY for read the full text of article
- Buy
- 500.00 rub