AutoML Approach to Stock Keeping Units Segmentation

Abstract

A typical retailer carries 10,000 stock-keeping units (SKUs). However, these numbers may exceed hundreds of millions for giants such as Walmart and Amazon. Besides the volume, SKU data can also be high-dimensional, which means that SKUs can be segmented on the basis of various attributes. Given the data volumes and the multitude of potentially important dimensions to consider, it becomes computationally impossible to individually manage each SKU. Even though the application of clustering for SKU segmentation is common, previous studies do not address the problem of parametrization and model finetuning, which may be extremely tedious and time-consuming in real-world applications. Our work closes the research gap by proposing a solution that leverages automated machine learning for the automated cluster analysis of SKUs. The proposed framework for automated SKU segmentation incorporates minibatch K-means clustering, principal component analysis, and grid search for parameter tuning. It operates on top of the Apache Parquet file format, an efficient, structured, compressed, column-oriented, and big-data-friendly format. The proposed solution was tested on the basis of a real-world dataset that contained data at the pallet level.