Recommendation Engine Built on the Most Innovative Algorithms

Unlike traditional rule-based personalization systems, Recombee's AI driven solution reflects real-time changes and complexity of user behavior online and thus enables you to personalize 1:1 improving your user experience and KPI's. Our solution analyzes user interactions and behavior online as well as product attributes and generates recommendations which are more likely to spark the interest of the customer.

At Recombee, we conduct research in collaboration with academia, which keeps us at the leading edge of innovation. We utilize deep-learning and collaborative filtering, as well as content-based algorithms (such as image and text processing algorithms) to ensure the most accurate content for all visitors. What is more, our solution is real-time, meaning that our models are adapting after every interaction of the users.

Collaborative Filtering

Models based on behavioral patterns

First type of the models Recombee adds to its model ensembles are Collaborative Filtering models, which are built from collected user-item interactions, such as detail-views or purchases. By analyzing behavioral patterns across the whole userbase, the recommendations are based on extracting interactions similarities between users, items, or both. “Similar users also liked” or “others also purchased” are both examples of CF-based recommendations. Recombee uses following CF models: Matrix Factorization, Nearest Neighbor methods, and Association Rules.

Content-Based

Models based on attributes of items and users

We also use Content-Based models, which estimate similarities between items or users by analyzing the provided property values. For example, two items can be considered similar by having similar categorization, name, text descriptions, etc. Various models are used to process different type of attribute data. These models are especially useful in cold - start situations when there’s not enough interaction data yet(brand new item or user).

Deep Learning

Models combining interaction and attribute data together

Following the cutting-edge research in the field, Recombee offers models based on neural networks and deep autoencoders to build the recommendations. Such models are able to consider at once all the data provided in the given context. The models build AI-based understanding of concepts hidden in the data.

Specialized Models

Models reflecting specific business-cases and product needs

Diversification models (recommending variety of different items), popularity-based models (long-term or trending), reminder models or periodicity-based models (based on repeating behavior in user-item interactions), are also part of Recombee. These come from vast amount of experience that Recombee team gained during years in business, applying the systems to hundreds of different use-cases.

Image Processing

Models based on analyzing images and visual similarity

Recombee can process product images to extract similarities based on visual style. This allows e.g. recommending items which are visually similar to those liked by a user in the past. Advanced models based on top of convolutional neural networks are used for that. Multiple images (such as photos taken from different angles) can be provided to further improve the performance.

AI-Based Model Optimization

Automated searching for proper hyperparametrization

The models in production need to continuously adapt to the changes in the environment such as different seasons or holidays. When put to production, the models adapt based on collected feedback. Knowing whether the recommendations really led to user actions allows Recombee to tune both the structure and the hyperparameters of the deployed model ensemble.

Large Language Models

Models based on semantic understanding

Another type of model included in Recombee's ensembles are Large Language Models (LLMs), which excel in understanding language and semantics. These models can capture subtle patterns and relationships in text, providing more accurate and context-aware recommendations. This capability is particularly valuable in applications like Semantic Search, where LLMs enhance the relevance of search results by understanding both user intent and the content of items. Recombee leverages various LLM approaches to further enrich its recommendation capabilities.