D.1: AutoML
• Short Description: AutoML is a software application that enables users to use and apply artifical intelligence (machine learning/deep learning), even if the users are not experienced in these topics. With a wizard like interface, AutoML provides abilities to define/select/clean data set, conduct pre-processing on the selected data set, perform data analysis, train, and test models based on provided algorithm set. By means of graphical interfaces, the application also compares the trained models based on different variables, such as F1 score, accuracy, precision, recall. The generated models are saved to be used later.
AutoML will be used in CircularTwAIN to cover all of the above listed functionalities. AI Module that will be used to reduce CO2 output from the sensors that are available on the plant and laboratory analysis as a label. This module includes data processing and helps to train multiple ML algorithms at once to compare results.
• Reference, URL: Avaliable upon request
• Relevant Domain/Industry: All/Smart Industry
• Application in relevant Projects/Initiatives: COGNITWIN
• Type: Web application (platform independent) It has a preconfigured pipeline architecture following wizard like steps implemented by functionalies in the order of provided tabs (data selection, preprocessing, analysis, algorithm selection, model training, model comparison, test, saving). LSTM, KNN, RF, GBT, MLP, SVM
• AI Breadth: ML, DL
• Learning Ability: Supervised Learning, currently, others can be added later
• Related technologies: Python, Scikit-learn, Keras, Flask (API), Java Script, HTML, CSS,React.js, PostgreSQL, TimeScaleDB
• License Information: Restricted
• Related to circularity and sustainability: RUL estimation and predictive maintenance models developed before resulted in energy minimization, and hence carbon footprint reduction. This AI module will be used to minimize steam consumption and CO2 output.
• Audience: Users, developers, industry independent
Go back to the AI portfolio of reference implementations
