mobile network traffic prediction based on user behavior and machine learning

One of the challenges of any network is user traffic analysis and prediction. mobile networks have achieved notable growth in recent years. traffic information on these networks play a crucial role in service quality control, user access control, and optimization. there are various methods for traffic prediction, including harmonic analysis and mathematical transformations, time series methods, and machine learning. due to the increasing volume of user data on mobile networks, machine learning methods have gained popularity in recent decades. in this paper, we introduce a probabilistic behavioral model and propose a new method that focuses on human behavior by categorizing users through clustering and utilizing their similarities. assuming a history of past user data is available, users with similar behavior are grouped into categories using clustering methods, and each category is assigned a label. the average of each cluster represents the traffic in that category. to predict the traffic of new users, our proposed method utilizes classification functions to determine the most appropriate category. subsequently, weighted averages are used to calculate the overall network traffic. we compare our proposed method with time series and fourier transform methods through three different scenarios. the results indicate that our method exhibits significant superiority over the other methods.