Dangdang search / recommendation technology leader Wang Hongtao, from the product point of view, personalized recommendation has penetrated into all aspects of Dangdang shopping process: shopping, shopping, shopping.
1, commodity information page
bought also bought see * * * * * * * *, * * * * at * * * * – based on a certain period of time order and access log data combined with a recommendation to buy orders based on data.
2, shopping cart,
will be based on the current user selected products for real-time computing, recommended goods.
3, Email mail
we will be in the order confirmation and feedback letters calculated according to the package with the commodity recommendation but in order to avoid the user to cancel the order to choose other goods, we are recommended to the user after the receipt of goods.
4, independent personalized page
collection of all data sources recommended Daquan, but at present the product data performance is not good, we are prepared to more in-depth integration with the shopping process. We also have to recommend that interest and potential mining congenial crowd, follow for its recent shopping behavior, extended recommendation. From the technical point of view, we mainly use content based (Dangdang commodity search is also responsible for our team), collaborative filtering, association rules and other methods. In the actual development, we also encountered cold start problems, especially we hope to promote the sales of merchandise commodities (the Department is weak, we will category) some categories of attributes of goods based on the recommendation of the size but not based on the product based on class.
our experience shows that personalized recommendation technology is an important means of e-commerce sites (especially large-scale e-commerce) (of course, can not be compared with the search). Rough estimate, the contribution to sales 10%~15% is no problem, which is the most important applications:
1, the information on the page to buy a * * * also bought.
2, product information on the page to buy recommended packaging.
3, shopping cart recommendation.