Implement personalized recommendations based on the negative feedback from usersLearn More >
Create a scene in the AIRec console and customize item selection and launching policiesLearn More >
Optimize diversified configurations based on business scenarios to improve user experienceLearn More >
Get a $50 voucher for your first purchase
Only for AIRec
End-to-End High-Quality Personalized Recommendation Service
Powered by Alibaba’s leading big data and artificial intelligence technologies, Alibaba Cloud Artificial Intelligence Recommendation (AIRec) provides personalized recommendation services for enterprises and developers. It is built on service accumulations in a variety of industries, such as e-commerce, content, news, livestreaming, and social media.
AIRec helps you associate users with items and content to achieve efficient algorithm-based recommendations.
Real-Time Recommendation Experience
Captures and analyzes user behavior in seconds and provides personalized recommendations within milliseconds. AIRec supports more than 1,000 QPSs. Users can refresh the page to view the latest recommended content.
Industry-Specific Algorithm Templates
Provides industry-specific algorithm templates and business-specific item selection and launching policies for scene recommendation. You can use algorithm models to implement fine-grained configurations to suit your business requirements.
High-Performance Cold Start
Improves user experience based on industry templates and real-time positive and negative feedback. This functions when no historical behavioral data is available in the early stage of the service release.
Personalized Recommendations on E-Commerce Platforms
Personalized recommendations have been gradually developed into an essential feature that increases the Gross Merchandise Volume (GMV) of e-commerce platforms.
Scenario Description: Users have differentiated short and long-term shopping interests and habits. They hope to find an assortment of items they like while shopping online and obtain a bonus. They also want to shop around on different platforms and check the prices of items they plan to purchase.
AIRec can be used in the "you may also like" and "related recommendations" scenarios to implement personalized recommendations.
Personalized Recommendations in PGC and UGC Communities
AIRec can quickly and accurately distribute the expected content from a pool that contains massive content.
Scenario Description: A large amount of content is generated every day. When users read content to learn knowledge, they can perform various actions to express their preference or understanding of the content, such as like, favorite, forward, and comment. Users can also perform an action, such as unlike or dislike, to provide negative feedback.
AIRec significantly improves the distribution efficiency of recommended content based on user interests and enhances on-site user engagement.
Personalized Recommendations for News
Major news and information platforms have embarked on a transformation towards efficient information distribution and user retention since the development of the Internet. AIRec can accurately distribute and deliver information amid rapid news updates.
Scenario Description: A large number of news entries are published (or expire) every day. Hot issues are updated in real-time. This requires AIRec to identify the potential news recommendations and accurately distribute news based on user interests. This way, users can enjoy personalized information services at higher reading efficiency.
Ease of Use
An end-to-end solution to stream processing that provides O&M and management. You can create an AIRec instance in a few hours.
Provisioning of End-to-End Features
Provides a series of auxiliary features throughout data interaction, testing, debugging, effect observation, and stable use. No extra development is required.
In-Depth Customization of Industry and Scene Templates
Provides differentiated, well-suited industry models and strategies by leveraging the rich experience of Alibaba in various industries and scenarios.
One-Step Personalized Recommendations
Delivers excellent user experience and provides real-time recommendations on Taobao’s homepage based on industry adaptation and upgraded recommendation algorithms.
Excellent Recommendation Effect and Performance
AIRec has proven its excellent recommendations and performance over many Double 11 promotional events. During these events, the AIRec architecture and algorithm models are used to handle high concurrent queries and large amounts of behavioral data.
Excellent Initial Effect
Regularly iterates industry templates to ensure the initial effect on the use of AIRec and user experience. AIRec can provide a recommendation performance that is 20% to 100% higher than self-managed algorithms.
Provides a variety of features based on years of business operations experience and combines operations logic with recommendation algorithms to open platform capabilities and implement rapid business customization.
High Security and Reliability
Isolates user data, encrypts sensitive information, delivers stable and reliable services, and responds to requests in milliseconds.
AIRec uses an offline or real-time data training model to model data based on short and long-term user interests and implements business-specific personalized recommendations.
How to Deploy AIRec
Please prepare business data, item data, and behavioral data for model training based on industry-specific data specifications
Design a Recommendation Page
Please design a recommendation page, rules to recommend items on the page, and user experience rules based on business requirements
Create a Recommendation Scene
Please create an instance, a recommendation scene based on the design of the recommendation page, and report offline and real-time data
Use a Fraction of Traffic to Verify the Recommendation Effect
Please use a fraction of traffic to verify the recommendation effect in the aspects of user experience and metric data.
Customize and Release a Recommendation Service
Please configure operations policies based on user experience requirements, use user traffic to conduct canary release, and then release the custom recommendation service