AWS Machine Learning Blog described the second part of the no-code machine learning scenario with Snowflake and Amazon SageMaker Canvas. In it, transactional data are connected to Canvas, prepared and merged using visual transformations with Data Wrangler, after which an XGBoost model is trained to detect fraud.
The material is devoted specifically to the stages of data preparation and model building. According to AWS Machine Learning Blog, the entire process is performed without writing code for machine learning and creates the basis for interactive dashboards in the third part of the series.
The practical significance of the approach lies in combining several stages of data work in a visual process. However the provided source contains only an editorial synopsis: it does not include model quality metrics, descriptions of the dataset, testing conditions, or independent confirmation of results.
editorial commentary
Why it matters
A probable consequence — further spread of visual workflows for fraud detection tasks among teams without deep programming skills. The next observable signal will be the third part of the series with interactive panels. Significant uncertainty remains around the quality of the model, data composition and its suitability for real operations.