Segmented Telecom Data Processing for Faster Event Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for detecting events in telecommunications networks, particularly in high-speed streaming networks, are inaccurate and inefficient due to their inability to effectively analyze long sequences of data items, leading to poor performance in capturing long-term temporal dependencies and slow prediction times.
Innovation Solution
A method involving segmenting a sequence of data items into segments and analyzing each segment using a machine learning model to identify relationships between data items, outputting a representation based on these relationships, which captures local information regardless of sequence length, allowing for more accurate and efficient event detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional machine learning models (RNNs, LSTMs, GRUs) are used to analyze long sequences of data items, then the model can capture temporal dependencies, but the prediction time becomes slow and the system becomes inefficient
Solution Approach 1:
The patent segments the long sequence of data items into multiple smaller segments or windows. Each segment is processed independently by the machine learning model, which significantly reduces the computational complexity and prediction time compared to processing the entire long sequence at once, while still capturing relevant temporal dependencies within each segment
Solution Approach 2:
Instead of analyzing the entire long sequence to detect events, the patent applies partial action by focusing analysis on smaller segments. This allows the system to detect events efficiently by examining only the relevant portions of the data sequence at any given time, improving prediction speed without sacrificing detection accuracy
2Measurement precision
If the sequence length increases to capture long-term temporal dependencies, then the model can detect rare events more accurately, but the computational complexity and resource requirements increase
Solution Approach 1:
The patent divides long sequences into smaller segments, reducing the computational complexity of processing each segment while maintaining the ability to detect rare events through appropriate segment design and aggregation of segment-level predictions
Solution Approach 2:
The patent transforms the problem from analyzing one long sequence in the time dimension to analyzing multiple shorter sequences in both time and segment dimensions. This dimensional transformation reduces computational complexity while preserving detection accuracy through multi-segment analysis
3Reliability
If traditional machine learning techniques are used on large volumes of streaming data, then comprehensive analysis can be performed, but the system becomes inefficient and cannot keep up with high-speed data streams
Solution Approach 1:
The patent segments the continuous streaming data into discrete windows or segments that can be processed independently and in parallel, enabling the system to keep up with high-speed data streams while performing comprehensive analysis within each segment
Solution Approach 2:
The patent applies periodic action by processing data in regular time intervals or fixed-size windows. This periodic processing approach allows the system to maintain comprehensive analysis while achieving the processing throughput required for high-speed streaming networks
Data Source
AI summary
There is provided a computer-implemented method for processing a sequence of data items. The data items correspond to one or more features of a telecommunications network. The sequence of data items is segmented into a plurality of segments. For each segment of the plurality of segments, the data items in the segment are analysed using a machine learning model to identify a relationship between the data items in the segment and output a representation of the data items in the segment based on the identified relationship.