Automated Feature Engineering for Time Series Data

Feature engineering for time series data, a critical task in data science, involves the transformation or encoding of raw data to create more predictive input features.This paper introduces a novel web framework designed to automate the labor-intensive and expertise-demanding process of time series feature engineering. The framework comprises advanced methods for automated feature extraction and selection, providing a wide range of application possibilities. A Bayesian Optimization strategy is also integrated to identify optimal features and model parameters for specific datasets, thereby enhancing prediction performance. The paper thoroughly explores the framework's design principles and operational procedures, along with validation of its effectiveness across different domains using real-world datasets.