Deep Learning Model for Real-Time Request Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Feature engineering in machine learning algorithms for detecting unauthorized activities is computation-intensive, slow, and requires significant storage, limiting configuration and operation, and often fails to leverage the sequence nature of data effectively.
Innovation Solution
A system that processes real-time requests by extracting a subset of variables, generating sequences from historical data, encoding string sequences, and inputting them into a trained deep machine learning model to compute a score indicative of unauthorized classification, leveraging advanced neural networks like transformer models for improved accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional feature engineering is used to detect unauthorized activities, then the system can identify suspicious patterns, but the computational intensity and processing time increase significantly
Solution Approach 1:
The patent replaces traditional mechanical feature engineering processes with a deep learning model that automatically processes raw data sequences. The transformer-based model substitutes manual feature extraction and engineering with automated neural network processing, achieving both high accuracy and improved processing speed through parallel computation and optimized architecture.
Solution Approach 2:
The patent changes the fundamental parameters of the detection system by transitioning from hand-crafted features to learned representations. The deep learning model learns optimal feature representations directly from raw data, changing the parameter space from manually defined features to dynamically learned embeddings, which improves both accuracy and processing efficiency.
2Reliability
If traditional feature engineering is used, then the system can process requests, but the storage requirements and computational resources increase significantly
Solution Approach 1:
The patent extracts only the essential information needed for detection by processing raw request data through the transformer model. Instead of storing and processing all raw features separately, the model extracts and processes only the relevant patterns and sequences, reducing storage requirements while maintaining detection capability.
Solution Approach 2:
The patent merges multiple feature extraction steps into a single unified deep learning model. By combining sequence processing, feature learning, and pattern recognition into one integrated transformer architecture, the system reduces the total quantity of intermediate data and storage requirements compared to traditional multi-step feature engineering pipelines.
3Reliability
If traditional machine learning algorithms are used, then the system can detect unauthorized activities, but the configuration and operation flexibility is limited
Solution Approach 1:
The patent introduces dynamics into the detection system by using a transformer model that can adapt to different detection scenarios. The model's attention mechanisms and learnable parameters allow it to dynamically adjust to various unauthorized activity patterns, providing configuration flexibility and adaptability across different use cases while maintaining high detection accuracy.
Solution Approach 2:
The patent creates a universal deep learning model that can perform multiple detection functions. The transformer architecture is designed to handle various types of unauthorized activities through a single unified framework, enabling the system to detect different patterns and anomalies across diverse scenarios without requiring separate specialized algorithms for each case.
Data Source
AI summary
A system and method include receiving a first set of variables associated with a real-time request, extracting a predetermined subset of the first set of variables for generating a second set of variables, identifying historical request data, computing a set of parameters based on the first set of variables and the historical request data, generating a plurality of numeric sequences and a plurality of string sequences for the real-time request, converting each of the plurality of string sequences into an encoded string sequence to obtain a plurality of encoded string sequences, inputting the plurality of numeric sequences and the plurality of encoded string sequences into a trained deep machine learning model, and computing a score from the trained deep machine learning model, the score indicative of a likelihood that the real-time request belongs to an unauthorized classification.