Cloud Asset Tagging via Hybrid TF-IDF and Deep Learning
Solution Overview
Problem
The rapid adoption of cloud computing by businesses poses challenges in managing costs, chargeback, capacity, and resource allocation due to unplanned expansion, and the manual classification of cloud computing assets becomes impractical as the number of assets grows, leading to a significant percentage remaining uncategorized.
Innovation Solution
A method and system for feature generation in asset classification using a combination of machine learning models, including deep learning and modified TF-IDF, to extract candidate features and predict tag information for cloud computing assets, allowing for accurate and explainable classification and prioritization of accuracy or explainability based on user preference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual classification of cloud computing assets is used, then classification accuracy may be maintained for small numbers of assets, but productivity deteriorates significantly as the number of assets grows
Solution Approach 1:
The system enables automated self-service classification of cloud assets through machine learning models that automatically process asset attributes, generate features, and predict classification labels without human intervention, thereby maintaining high productivity while ensuring accurate classification
Solution Approach 2:
The patent replaces manual mechanical classification processes with automated machine learning systems that use deep learning models and modified TF-IDF algorithms to process asset data, substituting human labor with computational intelligence to achieve both high speed and accuracy
2Device complexity
If traditional TF-IDF feature extraction is used, then computational simplicity is maintained, but measurement precision of feature representation deteriorates for complex cloud asset data
Solution Approach 1:
The system combines multiple feature extraction approaches into a composite model that integrates traditional TF-IDF methods with deep learning techniques, creating a hybrid feature representation that leverages the strengths of both simple and complex methods to achieve superior classification performance
Solution Approach 2:
The patent merges traditional TF-IDF feature extraction with deep learning-based feature generation, combining the computational efficiency of statistical methods with the representational power of neural networks to create a unified feature extraction pipeline
3Measurement precision
If deep learning models are used for feature extraction, then measurement precision of feature representation is improved, but device complexity increases
Solution Approach 1:
The system segments the feature extraction process into distinct components: traditional TF-IDF feature extraction, deep learning feature generation, and feature selection modules, allowing each component to be optimized independently and managed separately to reduce overall system complexity
4Productivity
If automated machine learning classification is implemented, then productivity is improved for large numbers of assets, but measurement precision may deteriorate without proper feature engineering
Solution Approach 1:
The system performs preliminary feature extraction and feature selection before the main classification process, preparing high-quality input features in advance to ensure that the automated classification maintains high accuracy while processing large volumes of assets efficiently
Data Source
AI summary
An embodiment includes generating an input document that includes a plurality of text fields of attribute data. The embodiment also includes extracting a set of candidate features from the attribute data using a feature extraction module that evaluates the attribute data using first and second machine learning models, where the first machine learning model scores terms in the input document and the second machine learning model includes a deep learning model. The embodiment also includes calculating feature-selection values for respective features of the set of candidate features and generating tag information for a remote computing asset using a machine learning classifier that predicts the tag information based on the feature-selection values.