Deep Architecture on an Immense Scale for the Purpose of Providing Customized Grocery Basket Recommendation

Abstract

In light of the growing number of consumers purchasing products online through platforms such as Instacart and Amazon Fresh, it is imperative that companies offer pertinent recommendations throughout the entire customer experience. This paper presents RTT2Vec, a supply chain-based grocery recommendation system. Its purpose is to provide accurate and tailored product suggestions in real-time to enhance the user's existing shopping cart. A comprehensive offline analysis of our system reveals a 8.4% enhancement in prediction metrics when contrasted with the state-of-the-art within-basket recommendation methods that serve as the baseline. Additionally, we offer an approximate inference method that is 12.4 times more efficient than exact inferred approaches. Our technology has enabled consumers to expedite the purchasing process, improved product discovery, and increased the mean container size as a result of its deployment.