Article Preview
TopA variety of tools designed to assist in machine learning processes have emerged over time. For this research, we identified and focused on three distinct types of tools. The first category includes tools for developers and data scientists, providing comprehensive programming libraries to facilitate the development of machine learning applications. TensorFlow, for example, is a machine learning framework that runs at scale and in diverse contexts (Abadi et al., 2016), assisting academics in pushing state-of-the-art models in ML and developers in simply building and deploy (ml)ing ML-powered apps. Apache Mahout, a library for scalable machine learning on distributed dataflow systems, is another example (Anil et al., 2020). In this category, we may also add Python libraries like PyTorch, Scikit-learn, or Keras.io, as well as cloud services like Google Colab, which is a serverless Jupyter notebook environment (Bisong, 2019).
Second, there are systems aimed at specialists while still providing tools for non-specialist users. These applications provide visual environments that aid in the visual development of machine learning models. Weka, for example, is a library of machine-learning techniques for data mining jobs. It features four environments, the most notable of which is Knowledge Flow, a visual interface that allows users to describe a data stream by visually linking components representing data sources, preprocessing tools, learning algorithms, assessment techniques, and visualization tools (Frank et al., 2009; Hall et al., 2020). RapidMiner Studio is a data science platform that includes tools for creating ML processes. It includes the Visual Workflow Designer tools for creating ML processes, and every step is documented for total transparency. This feature enables data source connection, automatic in-database processing, data visualization, and model validation processes (Bjaoui et al., 2020). KNIME Analytics Platform is another example. It gives tools for constructing visual data analytics workflows with a graphical interface that does not require scripting. KNIME is a modular platform that allows for the simple visual building and interactive execution of data pipelines (Berthold et al., 2009).
ML has also begun to be introduced in elementary and high schools. This has led to the creation of tools to assist non-expert users, such as children, in performing simple ML tasks using a visual interface. Two instruments can be highlighted in this category. The first is LearningML (Rodríguez-García et al., 2021), which is a platform for developing computational thinking abilities through hands-on AI projects, and second is Machine Learning for Kids (https://machinelearningforkids.co.uk/). Both tools are built around a primary pipeline for training models and a Scratch integration for using the trained model.
There are several sophisticated apps aimed at simplifying the use of ML algorithms, as well as instructional resources for comprehending these complex procedures. However, these platforms are designed to cater to a wide audience but often lack the specialized features necessary to address the unique requirements and challenges of specific domains. Against this backdrop, KoopaML emerges as a specialized visual ML platform that not only simplifies the creation and use of ML models for medical practitioners but also places a strong emphasis on the educational aspect of its use. It stands apart by integrating health-related criteria into its functionality, thus providing a more relevant and context-aware experience for users in the health domain. Moreover, KoopaML is dedicated to imparting an educational experience. It is structured to guide users through the underlying principles of ML techniques as they apply them, thereby fostering a deeper understanding of the rationale behind their choices. This contrasts with many existing solutions that may facilitate mechanical application of ML algorithms but fall short in educating users about the intricacies of model selection, data preprocessing, and algorithmic decision-making.