In the event of heavy rain last night, I participated in the theme speech entitled "Application of smart recommendation in mobile e-commerce" sponsored by it Longmen array. In his speech, Hou Xun proposed a "discovery-style unconscious shopping" idea. Many people may not go to a mall or press the road to buy anything, or to buy something, but during the shopping process, some promotion information, some eye-catching advertisements, and some interesting things are found, which can generate purchase behaviors.
Thanks to the mobility of mobile phones, they can be used everywhere and are everywhere. I think it makes more sense to implement "discovering unconscious shopping" on the mobile Internet. Shopping by users and shopping by users on mobile phones are all about user behavior, but how can we provide users with a virtual shopping environment, that is, this street, so that users can act with no purpose, this is a question that mobile Internet application developers should consider when they are interested in and promote consumption.
In real life, my shopping environment, that is, this street, is normal. If you go today or tomorrow, the commodities in the store generally do not change much. However, this situation does not exist in the virtual environment of the mobile Internet. It does not have time or culture. In reality, when you find a product that you are interested in the store, the clerk will explain it to you. Of course, there will be no similar roles on the network, but I think smart recommendations provide similar functions.
Intelligent Recommendation, recommendation engine, and personalized Intelligent Recommendation are all indispensable functions and research topics in the e-commerce field. Through the analysis of user behavior, User Classification and aggregation, we recommend products that users may be interested in. This is the directory of the product recommendation engine.
Personalized smart recommendations can also be called context-based recommendations. Personalized user recommendations are different from universal recommendations. When users use applications, there are some scenarios at the time, that is, context information, such as time, geographical location, activity status, and network status.
Context-based Personalized Intelligent Recommendation is a difficult topic currently being studied in the academic field. As far as I know, there is no complete model. In practical applications, I think this can be done. One method is to filter context information before data processing, and the other method is after data processing, the context information is used to filter the results. The third method is to participate in Data calculation based on the context information at runtime. I think the first two methods should be easier.
both traditional universal recommendations, personalized recommendations, and context-based Personalized recommendations are inseparable from basic mathematical theories. However, for actual application scenarios, use the algorithm that suits you. The model must be adjusted gradually in practical applications.