AI-based applications can help e-commerce retailers better understand who their customers are, what they like and dislike, and how they shop. This knowledge, in turn, offers endless ways to improve the customer experience. Retailers can create more personalized interactions by using AI to provide customers with unique recommendations for products to consider, products based on category preferences, reminders for replenishment, or curated items. Occasionally, a new customer does not have any previous purchase history. In this case, a product recommendation system can suggest products by analyzing the product descriptions from customer browsing history. Once a customer makes a purchase, the product recommendation system updates and recommends other products based on the purchase history and product ratings provided by other users.
This reference kit may help e-commerce retailers present targeted product recommendations to customers by applying textual clustering analysis to the product descriptions.
In collaboration with Accenture*, Intel developed this product recommendation AI reference kit. Paired with Intel® software, this kit may help to improve the performance of unsupervised learning models while generating product recommendations from product descriptions. This reference kit includes:
- Training data
- An open source, trained model
- User guides
- Intel® AI software products
At a Glance
- Industry: Retail
- Task: Recommend products based on the textual clustering analysis of the product descriptions
- Dataset: Dataset with textual product descriptions
- Type of Learning: Unsupervised learning
- Models: K-means clustering
- Output: List of product recommendations within the predicted cluster
- Intel® AI Software Portfolio:
- Intel® AI Analytics Toolkit (AI Kit)
- Intel® Extension for Scikit-learn*
- Intel® oneAPI Data Analytics Library (oneDAL)
Optimized with Intel® AI Software Products for Better Performance.
The product recommendation model was trained using Intel Extension for Scikit-learn. The Intel oneAPI Data Analytics Library (oneDAL) was used as the underlying solver to optimize the inferencing of the model.
Performance benchmark tests were run on Microsoft Azure* Standard_D8_v5 using 3rd generation Intel® Xeon® processors to optimize the kit.
To build a successful product recommendation system, data scientists need to train unsupervised learning models using substantial datasets. The ability to accelerate training allows the data scientists to train models more frequently and improve accuracy. Besides training, faster speed during inferencing allows retailers to provide product recommendations in real-time scenarios. This reference kit implementation provides performance-optimized guidance for product recommendation systems, which can be scaled across similar use cases that require K-means clustering.