Cellular Traffic Prediction Model Selection for Cloud Resource Balancing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Predicting resource utilization in cellular networks is challenging due to underutilization or overutilization, especially with the shift to cloud computing resources, which can lead to inefficiencies and resource waste.
Innovation Solution
A system and method for determining an optimum network prediction model by selecting and ranking various models (ARIMA, PROPHET, LSTM) based on target requirements such as accuracy, seasonality, and user ease of use, then employing the best model to predict traffic loads on target nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If cloud computing resources are dedicated for the cellular network, then network management functions can be moved to cloud resources, but resources may be underutilized or overutilized leading to waste or inefficiency
Solution Approach 1:
The system dynamically selects and switches between different prediction models (ARIMA, PROPHET, LSTM, etc.) based on current network conditions and performance metrics. This dynamic adaptation allows the system to optimize resource utilization in real-time, preventing both underutilization and overutilization of cloud computing resources while maintaining the flexibility benefits of cloud-based network management.
2Measurement precision
If multiple prediction models are evaluated and switched between, then prediction accuracy can be optimized, but system complexity increases
Solution Approach 1:
The system employs a lightweight model evaluation framework that quickly assesses multiple prediction models and discards underperforming ones in favor of better alternatives. Rather than maintaining complex permanent structures for all models, the system uses temporary, easily replaceable model evaluations that can be quickly switched based on current performance, reducing overall system complexity while maintaining high prediction accuracy.
Solution Approach 2:
The system changes key parameters such as prediction time horizons, data window sizes, and model selection criteria based on current network conditions. By dynamically adjusting these parameters rather than maintaining fixed complex model structures, the system achieves high prediction accuracy across varying traffic patterns while keeping the overall system architecture relatively simple and adaptable.
Data Source
AI summary
Systems and methods are directed towards determining an optimum network prediction model to employ to predict traffic loads on target nodes. A plurality of possible network prediction models and a plurality of target requirements in which to assess the plurality of possible network prediction models are selected. The plurality of target requirements are then ranked for the target nodes. For each corresponding possible network prediction model of the plurality of possible network prediction models, a total score is calculated based on scores of each corresponding target requirement and the rankings of each corresponding target requirement. An optimum network prediction model is determined from the plurality of possible network prediction models for the target nodes based on the total score of each of the plurality of possible network prediction models. The optimum network prediction model is then employed to predict the traffic loads on the target nodes in the cellular network.