Self-adaptive ANN Framework for Water Quality Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine-learning techniques, particularly ANN models, face challenges in predicting water quality due to limitations in capturing extreme or unexpected events and requiring manual adjustment of architectures and feature selection for different scenarios and geographical regions.
Innovation Solution
A self-adaptive optimization framework that automatically optimizes the ANN architecture and selects impact features using differential evolution (DE) and directional change (DC) based features, enabling the model to capture significant change patterns in water quality without increasing the sampling rate or requiring manual intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the sampling rate is increased to capture extreme or unexpected events in water quality data, then the ability to detect significant change patterns is improved, but the training cost and computational complexity of the machine-learning model increases
Solution Approach 1:
The patent extracts only the most relevant features from the raw water quality data using feature selection techniques. Instead of processing all raw data at high sampling rates, the system identifies and extracts key features that are most indicative of extreme events, thereby reducing the computational burden while maintaining detection capability.
Solution Approach 2:
The patent performs preliminary data processing and feature extraction before feeding data into the machine-learning model. By pre-processing the data to identify potential extreme events and extract relevant features in advance, the system reduces the computational complexity during model training while preserving the ability to detect significant change patterns.
2Measurement precision
If manual adjustment of ANN architecture and feature selection is performed by domain experts, then the model accuracy for specific scenarios is improved, but the adaptability to different geographical regions and situations deteriorates
Solution Approach 1:
The patent implements dynamic and adaptive feature selection that automatically adjusts to different geographical regions and water quality scenarios. Instead of using fixed manually-selected features, the system dynamically identifies relevant features based on the specific characteristics of each region, thereby maintaining high prediction accuracy across diverse environments without requiring manual reconfiguration.
Solution Approach 2:
The system performs self-service by automatically selecting features and optimizing ANN architecture without requiring manual intervention from domain experts. The feature selection process is automated through algorithms that identify relevant features based on data patterns, enabling the model to adapt to different geographical regions autonomously while maintaining high prediction accuracy.
3Measurement precision
If relevant features are carefully selected to avoid irrelevant noise, then the prediction performance is improved, but the complexity of feature selection process and manual intervention required increases
Solution Approach 1:
The patent performs preliminary feature selection and filtering before the main prediction process. By pre-identifying and selecting relevant features in advance using automated algorithms, the system reduces the complexity of the feature selection process while ensuring that only meaningful features are used for prediction, thereby improving prediction performance without requiring complex manual intervention.
Solution Approach 2:
The system automatically performs feature selection without requiring manual intervention from domain experts. The feature selection process is self-service, where algorithms autonomously identify and select relevant features based on data patterns and statistical analysis, thereby simplifying the overall process while maintaining high prediction performance.
4Measurement precision
If the ANN model is retrained for different situations or geographical regions, then the prediction accuracy for that specific region is improved, but the time and resources required for model deployment deteriorates
Solution Approach 1:
The patent performs preliminary feature selection and model configuration that is adaptable to different regions. By pre-processing the data and identifying region-specific features in advance, the system reduces the time required for model retraining when deploying to new geographical areas, as the framework is already prepared to handle different data characteristics.
Solution Approach 2:
The system uses dynamic feature selection that automatically adapts to different geographical regions without requiring complete model retraining. The feature selection process dynamically adjusts to regional characteristics, allowing the model to be deployed quickly in new areas while maintaining high prediction accuracy, thereby reducing the time and resources required for deployment.
Data Source
AI summary
The framework predicts water-quality data from observation data associated with raw features, e.g., rainfall. In the framework, an artificial neural network (ANN) computes the predicted water-quality data from feature data associated with impact features and derived from the observation data. The impact features are learnable, and are selected from impact-feature candidates comprising directional change- (DC-based) features each indicating occurrence of DC events in a corresponding raw feature. Including the DC-based features in the candidates enhances the ANN's ability of capturing significant change patterns of water quality due to extreme/unexpected events. The ANN architecture is also configurable according to model hyperparameters, which are learnable. The impact features and model hyperparameters are learnt by differential evolution for maximizing a prediction performance achieved by the ANN, thereby enabling the ANN architecture and impact features to be automatically optimized without requiring manual adjustment by domain experts in applying the ANN to different situations.