Tensor-Based Time Series Pattern Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for identifying patterns in time-series data are time-consuming, labor-intensive, and prone to human bias due to reliance on feature extraction methods that require extensive human input and are computationally costly, often resulting in information loss.
Innovation Solution
A system and method that transforms multivariate time-series data into tensors, trains a model using training data, and identifies behaviors using predictions from the trained model, eliminating the need for feature engineering and reducing human bias.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional feature extraction methods are used to identify patterns in time-series data, then pattern recognition can be performed, but the process becomes time-consuming and labor-intensive
Solution Approach 1:
The patent extracts and eliminates the feature extraction step from the conventional pattern recognition pipeline. By directly applying machine learning models to raw time-series data, it removes the time-consuming manual feature engineering process while maintaining pattern recognition capability through automated learning of relevant features from the data itself.
Solution Approach 2:
The machine learning model performs self-service by automatically learning and extracting relevant features from the raw time-series data without human intervention. The model adapts to the data characteristics and identifies patterns autonomously, replacing the manual feature extraction process that required domain expertise and significant time investment.
2Measurement precision
If conventional feature extraction methods are used, then patterns can be identified, but the process becomes computationally costly and requires losing information
Solution Approach 1:
The patent removes the intermediate feature extraction stage that caused information loss. By feeding raw time-series data directly into the machine learning model, it preserves all original data information including temporal relationships, correlations, and subtle patterns that would be lost during manual feature engineering and dimensionality reduction.
Solution Approach 2:
The patent changes the approach from manual parameter selection (feature engineering) to automated parameter learning through machine learning. The model automatically determines which features and parameters are most relevant for pattern recognition, adapting to the specific characteristics of the data without requiring pre-defined features or losing information through rigid feature transformations.
3Measurement precision
If conventional feature extraction relies on human engineering, then patterns can be identified, but human bias is introduced into the process
Solution Approach 1:
The patent implements self-service by replacing human engineers with machine learning algorithms that automatically learn features from data. The model objectively identifies patterns based on statistical relationships in the data without being influenced by human preconceptions, biases, or subjective judgments, thereby eliminating the harmful effect of human bias while maintaining pattern recognition capability.
4Measurement precision
If feature extraction is performed manually, then patterns can be identified, but the process becomes labor intensive
Solution Approach 1:
The patent extracts and eliminates the manual labor component from pattern recognition by removing the feature extraction step. Machine learning models automatically process raw data and identify patterns without human intervention, transforming a labor-intensive process into an automated computational task that significantly improves productivity and efficiency.
Solution Approach 2:
The patent substitutes the mechanical process of manual feature extraction with an automated machine learning system. Instead of human analysts manually examining and extracting features from time-series data, the system uses algorithms to automatically learn and identify patterns, replacing human labor with computational processes that are faster, more consistent, and scalable.
Data Source
AI summary
Systems and methods identify a behavior in multivariate time-series data. The systems and methods receive data representing a time series having steps each representing an event associated with a time stamp and being associated with one or more attributes; transform the data into a tensor; train a model using training data comprising a set of tensors; identify the behavior using predictions from the trained model and a target pattern; and provide an indication of the presence or absence of the behavior.