An AI-Driven Approach to Industrial Testing: From Time Series to Supervised Classification through Dynamic Thresholds

Dynamic Time Series Modeling with AI-Assisted Data Labeling is a pipeline for analyzing time-series data. It integrates deep learning and machine learning techniques to detect anomalies, dynamically label data, and classify observations. The workflow includes:

1. Deep LSTM Autoencoder: Extracts latent features from time-series data to capture temporal patterns.
2. Supervised Data Labeling: Uses Ridge Regression with dynamic thresholds to label windows.
3. Data Augmentation: Balances the dataset with SMOTE and under-sampling techniques.
4. CNN Classification: Classifies time-series windows into categories with high accuracy.
5. Evaluation and Visualization: Provides insights through regression plots, feature importance, and ROC curves.

System Architecture

This modular approach ensures scalability, interpretability, and robust performance for time-series modeling and classification.