Flash Call Detection Using CDR Patterns and ML Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing rule-based methods for detecting flash calls in mobile networks are inaccurate and inefficient, leading to increased costs and network burden for mobile network operators due to unidentified flash calls.
Innovation Solution
A machine learning-based approach using call detail records (CDRs) and call network data to identify flash calls by analyzing distinctive call patterns, employing features such as call duration, answer status, and CallerID behavior, and implementing intervention actions like call blocking, delay, or diversion to prevent flash calls.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If rule-based methods are used to detect flash calls, then the detection process is simple to implement, but the detection accuracy is low leading to unidentified flash calls
Solution Approach 1:
The patent replaces traditional rule-based mechanical detection systems with a machine learning-based intelligent system. The machine learning model analyzes multiple call features (duration, timing, frequency, CallerID patterns) to automatically distinguish flash calls from legitimate calls, achieving high detection accuracy without requiring complex manual rule configuration.
Solution Approach 2:
The patent transforms the detection approach by changing from fixed rule parameters to dynamic learned parameters. The machine learning model learns optimal detection parameters from training data, adapting to different flash call patterns and evolving threats, thereby improving accuracy while maintaining system simplicity.
2Loss of energy
If flash calls are not identified, then the detection system remains simple, but network costs increase due to high volume of missed calls
Solution Approach 1:
The patent implements preliminary classification of calls using a machine learning model trained on historical call data. By pre-identifying flash call patterns before they consume network resources, the system can block or flag these calls early in the call setup process, preventing unnecessary network resource consumption and reducing costs.
Solution Approach 2:
The system incorporates feedback mechanisms where detection results and network cost data are continuously fed back to retrain and improve the machine learning model. This closed-loop approach ensures the system learns from actual network conditions and optimizes its detection accuracy to minimize ongoing network costs.
3Productivity
If flash calls are allowed to proceed, then network activity increases providing service functionality, but network burden increases due to paging and signalling traffic
Solution Approach 1:
The patent extracts and isolates flash call traffic from legitimate network traffic through machine learning-based classification. By identifying and separating flash call patterns (such as immediate termination after ringing, specific CallerID behaviors), the system can apply targeted blocking or routing rules only to flash calls, allowing legitimate traffic to proceed unaffected while reducing overall network burden from flash call volume.
4Measurement precision
If machine learning model is implemented for flash call detection, then detection accuracy improves, but computational resources and processing time increase
Solution Approach 1:
The patent performs preliminary training of the machine learning model offline using historical call data, storing the trained model for rapid inference. During actual call detection, the pre-trained model quickly classifies calls based on learned patterns, minimizing real-time processing delays while maintaining high accuracy.
Solution Approach 2:
The system applies machine learning analysis selectively to calls with suspected flash call characteristics based on preliminary filtering rules. By not applying full machine learning analysis to every call but only to those meeting certain criteria, the system reduces overall processing time while maintaining high detection accuracy for flash calls.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
There is provided a method for training a machine learning model for identifying flash calls in a set of Call Detail Records, CDRs, the method comprising: receiving a set of CDRs and a set of call network data; creating a first training set, the first training set comprising a first subset of CDRs from the set of CDRs and a second subset of CDRs from the set of CDRs, wherein the first subset of CDRs comprises a plurality of CDRs known to represent flash calls, and the second subset of CDRs comprises a plurality of CDRs known to represent legitimate calls; determining one or more characteristic features in the first training set, the one or more characteristic features comprising a first characteristic feature associated with the first subset of CDRs and a second characteristic feature associated with the second subset of CDRs, wherein the first characteristic feature is different from the second characteristic feature.