Encoded State-Change Clustering for Future Outcome Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Analyzing past system state instances in complex, multi-dimensional stateful systems with large volumes of transactions to identify relationships with future system states is computationally overwhelming due to the volume of data and fleeting changes, making traditional analysis impractical.
Innovation Solution
Cluster unique machine-learned encoded historical system state changes based on probable relationships with subsequent outcomes, using an LSTM autoencoder to reduce data volume, and then cluster by similarity to outcomes within a threshold period, enabling efficient search and analysis of past system states related to target outcomes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional analysis methods are used to review large volumes of system state data, then comprehensive analysis coverage is achieved, but computational complexity becomes overwhelming
Solution Approach 1:
The patent segments the large volume of system state data into individual state instances, each representing a discrete change in system state. By processing and analyzing these segmented state instances separately through machine learning encoding, the system achieves comprehensive analysis coverage while avoiding the computational overwhelm of processing all data simultaneously as a monolithic block.
2Measurement precision
If all historical system state data is processed to identify relationships with future states, then prediction accuracy is improved, but processing time increases significantly
Solution Approach 1:
The patent applies preliminary action by using machine learning models to encode and pre-process historical system state data into compressed representations before actual analysis is needed. This preliminary encoding transforms raw state data into a more manageable format, enabling faster subsequent processing and relationship identification while maintaining prediction accuracy.
Solution Approach 2:
The patent changes parameters by transforming the representation of system state data through machine learning encoding. The raw high-dimensional state data is converted into encoded representations with different dimensionalities and features, allowing the system to identify relationships more efficiently while preserving the essential information needed for accurate predictions.
3Measurement precision
If detailed analysis of fleeting changes in system states is performed, then identification of causal relationships is improved, but data processing overhead becomes impractical
Solution Approach 1:
The patent introduces machine learning encoders as intermediary components between the raw system state data and the analysis process. These intermediaries process and transform the detailed fleeting changes into encoded representations that retain the essential relationship information while reducing the processing overhead, making causal relationship identification practical and efficient.
Data Source
AI summary
The disclosed embodiments relate to reducing a computational burden for identifying and analyzing past system state instances of complex, i.e., multi-dimensional or multi-variate, stateful systems which process large volumes of arbitrary or pseudo arbitrary transactions which modify the state thereof, where a prior system state instance may have an effect on a future system state instance, in order to, for example, discern some insight about actual or potential later occurring state instances. The disclosed embodiments cluster unique machine learnt encoded historical system state changes based on a probable/predictive relationship with one or more defined outcomes, each comprising one or more subsequent system state changes indicative thereof, forming an efficient outcome searchable database.