Supervised Machine Learning

Supervised Learning is the most common ML variant in production applications. The principle is simple: a human labels examples (this product belongs in category X, this data point is an anomaly), the model learns the patterns and applies them to new, unlabeled data. The quality of the model depends directly on the quality and volume of training data.

The two main application types are classification and regression. Classification assigns inputs to one of several categories (spam or not, product category A, B, or C, review positive or negative). Regression predicts a numerical value (price, inventory level, conversion probability). Both use similar algorithms with different output structures.

Pimcore uses supervised learning in several function areas. Product classification into eCl@ss/ETIM structures is trained on historical classification decisions, data quality checks learn from flagged anomalies, AI enrichment learns from enriched products. Through the Agent SDK, custom supervised learning models can be embedded for application-specific classifications.