Explainable Ensemble Feature Selection for Industrial Forecasting

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional feature selection methods struggle to handle complex and diverse industrial data, fail to capture nonlinear relationships, and lack interpretability and adaptability, leading to suboptimal performance in predictive models and hindering AI adoption in critical industrial applications.

Innovation Solution

A recursive ensemble feature selection method using explainable artificial intelligence (XAI) that includes data preprocessing, machine learning model training, feature contribution scoring, and iterative refinement to identify and eliminate less important features, ensuring transparency and adaptability.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional feature selection methods are used, then the process is simple, but predictive accuracy and handling of complex data deteriorate

Engineering Contradiction:
Improvepredictive accuracyVSAvoidfeature selection process complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The feature selection process is segmented into multiple iterative steps: initial feature selection using SHAP values, intermediate selection using LIME explanations, and final selection using DeepLIFT attributions. Each step handles a specific aspect of feature importance assessment, allowing the system to manage complexity through structured decomposition rather than monolithic processing.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a new dimension to feature selection by integrating explainable AI techniques (SHAP, LIME, DeepLIFT) alongside traditional methods. This multi-dimensional approach assesses feature importance from multiple perspectives simultaneously, improving predictive accuracy without relying on a single simplified method.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Measurement precision

If AI models are used to improve predictive performance, then accuracy improves, but interpretability and transparency deteriorate

Engineering Contradiction:
Improvepredictive accuracyVSAvoidinterpretability
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent uses explainable AI techniques (SHAP, LIME, DeepLIFT) as intermediary layers between the complex AI models and the decision-making process. These intermediaries translate model predictions into human-understandable explanations, preserving both predictive accuracy and interpretability by acting as a bridge between the black box model and transparent decision-making.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system implements feedback mechanisms where feature importance scores from explainable AI techniques are used to refine the feature selection process, which in turn improves model performance. This iterative feedback loop ensures that both accuracy and interpretability are maintained, as the model continuously learns from explained patterns.

Inventive Principle:
Principle #23Feedback

3Adaptability or versatility

If traditional modeling techniques are used, then the system is simple, but ability to capture nonlinear relationships and adapt to dynamic environments deteriorates

Engineering Contradiction:
Improveadaptability to dynamic environmentsVSAvoidmodeling system complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements dynamic feature selection by continuously updating feature importance scores based on real-time data patterns and changing system conditions. The explainable AI techniques adapt to new data distributions and temporal patterns, allowing the system to dynamically adjust its feature selection strategy rather than relying on static pre-defined features.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system performs preliminary feature selection and grouping before final modeling, organizing features into temporal and spatial groups in advance. This preliminary structuring enables the model to better handle dynamic environments by pre-organizing data relationships that will emerge during operation.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4668170A1Method and system for recursive ensemble feature selection using explainable artificial intelligence
Publication Date: 2025.12.24 TATA CONSULTANCY SERVICES LTD
  • EP4668170A1 patent drawingFigure 1
  • EP4668170A1 patent drawingFigure 2
  • EP4668170A1 patent drawingFigure 3

AI summary

The embodiments of the present disclosure herein address unresolved limitations in handling feature selection, inadequacies in capturing nonlinear relationships, issues related to interpretability in Artificial Intelligence (AI) models, and a lack of adaptability to the dynamic nature of industrial environments. Embodiments herein provide a method and system for a recursive ensemble feature selection using an explainable artificial intelligence (XAI). The method begins with a thorough digital modelling process, where the careful selection of features and identification of optimal time lags take precedence. By utilizing XAI techniques, the system ensures a transparent and interpretable selection of features and time lags. This method forms a robust foundation for subsequent forecasting model development tailored to specific datasets. The infusion of XAI into the feature selection and time lag identification stages enhances the accuracy of forecasting models while providing clear insights into the factors influencing predictions.