IoT Data Analytics Ontology and ML Translation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current data analytics solutions in IoT platforms face challenges with inaccurate information cataloging across domains, inability to fine-tune hyperparameters and metrics for sensor data processing, and the complexity of interpreting domain-specific jargon, leading to inaccurate feature extraction and interpretation.

Innovation Solution

A system that employs a multi-layered abstraction framework with a cataloguing service supporting multiple views, uses ontology represented as a Directed Acyclic Graph (DAG) to capture stakeholder knowledge, applies machine learning techniques like Convolutional Neural Networks and tensor factorization to identify sensor relationships, and translates domain-specific jargon for cross-domain analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional data analysis and processing theories, technologies, tools, and processes are used, then the system is simple and easy to operate, but it cannot achieve in-depth understanding and discovery of actionable insights in big data

Engineering Contradiction:
Improvedata analytics accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary layer between traditional data processing and advanced analytics. This intermediary includes a cataloguing service that standardizes data from multiple domains, an ontology service that creates structured knowledge representations, and a machine learning service that performs automated feature extraction. These intermediaries transform raw sensor data into structured, interpretable formats that enable advanced analytics without requiring users to directly manage system complexity.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system segments the complex data analytics process into distinct, manageable services: data ingestion service, cataloguing service, ontology service, machine learning service, and visualization service. Each service handles a specific aspect of the analytics pipeline, allowing independent optimization and maintenance while collectively achieving advanced analytical capabilities.

Inventive Principle:
Principle #1Segmentation

2Loss of information

If information from across domains and subject matter experts is gathered, then more comprehensive knowledge is obtained, but it involves lots of jargon which are difficult to interpret and build relationship models

Engineering Contradiction:
Improveknowledge retentionVSAvoidinterpretability
Core Design Contradiction:
Loss of informationVSEase of operation

Solution Approach 1:

The ontology service transforms domain-specific jargon and unstructured expert knowledge into standardized parameters and structured relationship models. By converting qualitative expert insights into quantifiable ontological parameters with defined relationships, the system preserves comprehensive domain knowledge while making it computationally processable and interpretable across different domains.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system replaces manual interpretation of domain jargon with automated machine learning models. The machine learning service automatically extracts features and builds relationship models from structured ontological data, substituting human expert interpretation with algorithmic processing that can handle complex cross-domain relationships at scale.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If manual cataloging and interpretation of sensor data is performed, then domain expertise can be applied, but it is time-consuming and leads to inaccurate feature extraction from large volume data

Engineering Contradiction:
Improvefeature extraction accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The machine learning service implements self-service automated feature extraction that processes large volumes of sensor data without manual intervention. The system automatically ingests raw sensor data, applies learned models to extract relevant features, and generates insights autonomously. This self-service capability maintains high extraction accuracy by leveraging trained models while eliminating the time constraints of manual processing.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The cataloguing service performs preliminary organization and standardization of sensor data before it reaches the machine learning service. By pre-processing and structuring data in advance according to established ontologies, the system prepares data for efficient automated processing, reducing the computational burden and time required for feature extraction while maintaining accuracy.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3633959B1Automation of data analytics in an internet of things (IOT) platform
Publication Date: 2024.05.15 TATA CONSULTANCY SERVICES LTD
  • EP3633959B1 patent drawingFigure 1
  • EP3633959B1 patent drawingFigure 2

AI summary

Advanced analytics refers to theories, technologies, tools, and processes that enable an in-depth understanding and discovery of actionable insights in big data, wherein conventional systems and methods may be prone to errors leading to inaccuracies. Embodiments of the present disclosure provide systems and methods for performing data analytics in an IoT platform wherein input data pertaining to problem type, solution associated thereof, sensory information corresponding to sensors deployed in the platform, knowledge of domain expert(s) specific to input data are depicted in graphical knowledge representations, wherein relationships between (i) sensors and associated attributes and (ii) the domain knowledge across the plurality of graphical knowledge representations are determined wherein Machine Learning (ML) models are then applied to the determined relationship for optimizing ML models wherein problem statement, root cause, from one knowledge to another are translated using translation technique(s) to determine root cause analysis and interpretable root cause information.