Mixture-Exponential Model for Bursty User Activity in IPTV Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current models for user activity in network systems, such as IPTV, often fail to accurately represent real-world behavior, leading to incorrect system performance estimation due to oversimplification or lack of realistic data, especially in systems with bursty activity patterns.
Innovation Solution
A method is developed to model user activity by analyzing diurnal patterns and using a mixture-exponential model to estimate user activity information, which includes channel popularity, session lengths, and traffic demands, allowing for the modification of network parameters based on these estimates to improve system performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a constant-rate Poisson process is used to model user activity, then the model is simple and easy to implement, but it does not accurately model bursty activity patterns
Solution Approach 1:
The patent transitions from a constant-rate Poisson process to a non-homogeneous Poisson process with time-varying rate functions. The rate parameter λ(t) is changed from constant to a function that captures diurnal patterns and bursty behavior, allowing the model to adapt to changing user activity conditions while maintaining the Poisson process framework.
Solution Approach 2:
The patent introduces dynamic rate functions that vary over time to capture bursty user activity patterns. Instead of a static constant rate, the model uses time-dependent rate functions that reflect real-world user behavior patterns, including periods of high and low activity, making the model dynamic rather than static.
2Measurement precision
If actual trace data is used directly for system performance evaluation, then the modeling accuracy is high, but the data contains too much commercial information and user information to be publicly distributed
Solution Approach 1:
The patent creates synthetic copies of user activity data through simulation models that replicate the statistical properties and patterns of actual trace data without containing real user information. These synthetic datasets preserve the essential characteristics needed for system performance evaluation while being safe for public distribution and sharing.
Solution Approach 2:
The patent introduces simulation models as an intermediary between actual trace data and system performance evaluation. Instead of directly using sensitive real data, the models act as intermediaries that generate equivalent synthetic data, preserving evaluation accuracy while eliminating privacy and commercial sensitivity issues.
3Ease of operation
If previous user activity models are used, then the modeling process is straightforward, but the models are quite different from reality and can lead to incorrect estimation of system performance
Solution Approach 1:
The patent modifies the parameters of the Poisson process from constant rates to time-varying rate functions that capture diurnal patterns. This parameter change allows the model to reflect realistic user behavior patterns including daily cycles and bursty activity, significantly improving reliability while maintaining the mathematical tractability of the Poisson framework.
Solution Approach 2:
The patent incorporates periodic diurnal patterns into the rate functions to reflect the cyclical nature of user activity throughout the day. The rate functions include periodic components that capture morning and evening peaks, nighttime lows, and other recurring patterns, making the model more realistic without complicating the underlying Poisson process structure.
Data Source
AI summary
Systems and methods to model user activity information associated with a network system are provided. A particular method includes receiving, at a computing device, a request for user activity information associated with selected channels of a television access network that provides multimedia content to users. The method includes executing a model of user activity associated with the television access network at the computing device. The model estimates the user activity information as user multimedia access demands during particular time periods within a day. The method also includes storing the user activity information at a computer-readable non-transitory storage medium.