Network State Vector Prediction for Fault Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning algorithms, such as Deep Neural Networks and Recurrent Neural Networks (RNNs), struggle with predicting faults in mobile networks due to mismatched data structures, leading to reduced accuracy in proactive self-healing, as they are better suited for discrete data rather than continuous numerical values from network performance management data.
Innovation Solution
A network management apparatus that transforms continuous numerical network data into a state vector using vector quantization and distance transformation, enabling a self-learning prediction module, particularly an LSTM RNN, to predict future network states and facilitate proactive corrective actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If Recurrent Neural Networks (RNNs) are used to predict future network states from continuous numerical network data, then the system can perform proactive fault prediction, but the prediction accuracy degrades because RNNs are better suited for discrete data rather than continuous numerical values
Solution Approach 1:
The patent transforms the continuous numerical network data into discrete state vectors through vector quantization. This parameter transformation changes the data representation from continuous KPI values to discrete states that are appropriate for RNN processing, thereby resolving the mismatch between the algorithm's requirements and the data structure while maintaining prediction accuracy
2Ease of manufacture
If machine learning algorithms are applied to mobile network data without adapting to the unique environment, then implementation is simpler, but the algorithms do not utilize their fullest potential due to case-by-case differences in data structure and system behavior
Solution Approach 1:
The patent adapts the machine learning approach by transforming continuous network data into discrete state representations. This parameter transformation makes the data suitable for standard RNN algorithms while capturing the unique characteristics of mobile network data, including its temporal, spatial and dimensional correlations
Solution Approach 2:
The patent introduces vector quantization as an intermediary step between data collection and RNN processing. This intermediary transformation layer converts continuous KPI data into discrete state vectors, enabling the RNN to effectively process mobile network data while maintaining the benefits of standardized algorithms
Data Source
AI summary
Network management apparatus and methods are described. A network management apparatus comprises network data receiving means for receiving network data that is representative of the current condition of a communications network, the network data comprising a plurality of values indicative of the performance of network elements; network data transformation means for transforming the received network data into a network state vector that is indicative of a current state of the network; and network state prediction means for predicting a future network state vector of the network from the current network state vector, the network state prediction means comprising a self-learning prediction module having a memory for storing at least one internal state.