Like peanut butter? This algorithm has a hunc

Recommendation algorithms can make a customer’s online shopping experience faster and more efficient by suggesting complementary products each time the customer adds a product to their cart. Did the customer buy peanut butter? The algorithm recommends several jelly brands to add next.

These algorithms typically work by matching purchased items to items that other shoppers have frequently purchased alongside them. If the buyer’s habits, tastes, or interests closely resemble those of previous customers, such recommendations could save time, jog memories, and be a welcome addition to the shopping experience.

But what if the customer buys peanut butter to fill a dog toy or bait a mousetrap? What if the customer prefers honey or bananas with their peanut butter? The recommendation algorithm will offer less helpful suggestions, which will cost the retailer a sale and could annoy the customer.

New research led by Negin Entezari, who recently completed a PhD in computer science at UC Riverside, Instacart collaborators and his PhD supervisor Vagelis Papalexakis, brings a methodology called tensor decomposition – used by scientists to find patterns in massive volumes of data – in the world of commerce to recommend complementary products better suited to customer preferences.

Tensors can be represented as multi-dimensional cubes and are used to model and analyze data with many different components, called multi-aspect data. Data closely related to other data can be connected in a cube arrangement and related to other cubes to discover patterns in the data.

“Tensors can be used to represent customer buying behaviors,” Entezari said. “Each mode of a 3-mode tensor can capture one aspect of a transaction. Customers form one mode of the tensor, and the second and third modes capture product-to-product interactions by considering products co-purchased in a single transaction.

For example, three hypothetical buyers (A, B and C) make the following purchases:

A: Buy hot dogs, hot dog buns, coke and mustard in one transaction.
B: Makes three separate transactions: Cart 1: Hot dogs and hot dog buns; Basket 2: Coke; Basket 3: Mustard
C: Hot dogs, hot dog buns and mustard in one transaction.

For a conventional matrix-based algorithm, customer A is identical to customer B because they purchased the same items. Using the tensor decomposition, however, customer A is more closely related to customer C because their behavior was similar. The two had similar products purchased together in a single transaction, although their purchases differed slightly.

The typical recommendation algorithm makes predictions based on the item the customer just purchased, while the tensor decomposition can make recommendations based on what’s already in the entire basket of the user. So if a shopper has dog food and peanut butter in their cart but no bread, a tensor-based recommendation algorithm might suggest a refillable dog chew toy instead of jelly if other users also made this purchase.

“Tensors are multidimensional structures that allow modeling of complex and heterogeneous data,” said Papalexakis, associate professor of computer science and engineering. “Instead of just noticing which products are bought together, there is a third dimension. These products are purchased by this type of user and the algorithm tries to determine which types of users create this match. »

To test their method, Entezari, Papalexakis and co-authors Haixun Wang, Sharath Rao and Shishir Kumar Prasad, all researchers for Instacart, used a public Instacart dataset to train their algorithm. They found that their method outperformed state-of-the-art methods for predicting customer-specific complementary product recommendations. Although further work is needed, the authors conclude that big data tensor decomposition could find its place in large enterprises as well.

“Tensorial methods, while very powerful tools, are even more popular in academic research when it comes to recommender systems,” Papalexakis said. “For industry to adopt them, we need to demonstrate that it’s attractive and relatively easy to substitute for whatever they have that already works.”

While previous research has shown the benefits of tensor modeling in recommendation problems, the new publication is the first to do so within the framework of complementary element recommendation, bringing tensor methods closer to industrial adoption and commercialization. technology transfer in the context of recommender systems.

“Tensorial methods have already been successfully adopted by industry, chemometrics and food quality being prime examples, and each attempt like our work demonstrates the versatility of tensorial methods in being able to tackle such a wide range of challenging issues in different areas,” Papalexakis said.

The paper, “Tensor-Based Complementary Product Recommendation,” was presented at IEEE Big Data 2021 [2021 IEEE International Conference on Big Data] and is available here. The work, which began with Entezari’s internship project at Instacart, was partially funded by the National Science Foundation.

Warning: AAAS and EurekAlert! are not responsible for the accuracy of press releases posted on EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.

Source link

Comments are closed.