DeepTSF: Codeless machine learning operations for time series forecasting

This paper presents DeepTSF, a comprehensive machine learning operations (MLOps) framework aiming to innovate time series forecasting through workflow automation and codeless modeling. DeepTSF automates key aspects of the ML lifecycle, making it an ideal tool for data scientists and MLops engineers engaged in machine learning (ML) and deep learning (DL)-based forecasting. DeepTSF empowers users with a robust and user-friendly solution, while it is designed to seamlessly integrate with existing data analysis workflows, providing enhanced productivity and compatibility. The framework offers a front-end user interface (UI) suitable for data scientists, as well as other higher-level stakeholders, enabling comprehensive understanding through insightful visualizations and evaluation metrics. The application of DeepTSF in real-life use cases of the I-NERGY project has already proven DeepTSF's efficacy in DL-based load forecasting, showcasing its significant added value in the electrical power and energy systems domain. Historically, time-series modeling has been a prominent area of interest in academic research, with diverse applications in fields such as climate modeling [42], biological sciences [60], medicine [63], and commercial decision-making domains like retail [59], finance [56], and energy [49, 48]. Traditional approaches in this field have primarily focused on parametric statistical models, utilizing domain expertise-driven techniques such as autoregressive models [13], exponential smoothing [23], and other methods that heavily relied on decomposing time series [8]. However, the advent of modern ML methods has introduced data-driven approaches for capturing temporal dynamics [37]. Among these methods, deep learning (DL) has gained significant traction, inspired by its remarkable achievements in areas like image classification [31], natural language processing [66], and reinforcement learning [34]. Deep neural networks, with their customized architectural assumptions or inductive biases [10], can effectively learn intricate data representations, eliminating the need for manual feature engineering and model design. The availability of open-source backpropagation frameworks [46, 1] has further simplified network training, allowing for flexible customization of network components and loss functions.