Also Bought Machine Learning Recommendation Generator

A machine learning powered plugin for generating a recommendation feed based on purchase history. It aims to find implicit connections between products hence providing a more sophisticated customer journey than straightforward recommendation of products been bought together once. The widget appears on a product’s page and displays related products. The recommendations are ranked based on their likelihood being bought together in descending order. Technical details Building recommendations takes time and depends on your data and algorithm settings you choose. The richer your purchase history and products set the greater amount of time it takes to build recommendations. Once they are generated all data needed for the widget is stored in the database and no additional overhead is introduced when a product page is requested - it behaves as every other part of the application. After recommendations are built you can find some formal estimations of the model in logs. Important thing is that only you can tell if a recommendation is good enough. The first point is that recommendations should give the purchases that actually happened. For example if product A was bought with product B and C you should see them in the recommendation list. Then if product B was bought with D it is likely to find D in the recommendation list for A and so forth. After a model is generated you can find some metrics in logs to estimate its quality and tune model parameters. All data is being processed and stored on your site - it does not use any third party services nor sends any statistics or metrics.