all the recommendations are not playing tricks, not recommended, not recommended.
goes to the mall, you tell yourself you only buy T-Shirts today, and you still carry big bags when you’re out of the mall…… Shopping guide to see you touch the dress, let you try for free, you will not wear this off, and one after another. Shopping guide guess what customers like, recommend customers try on, to meet the psychological demands of both sides to reach a deal. Electricity supplier website how to guess the user’s mind, recommend commodities to reach a deal? We can’t make people aware of the user’s demands. We can only tell users by data and rules. I know you are looking at this product. All the recommendations do not tune is rogue, not recommended, not recommended.
1, recommendation algorithm
recommendation algorithms include content-based recommendation algorithm, collaborative filtering algorithm and population based statistical recommendation. First come under popular science these recommended algorithms:
1, the content recommendation algorithm (CB): extract feature modeling item for each
CB builds commodity model recommendation based on commodity correlation. Commodity relevance includes commodity categories, attributes, parameters, keywords, portfolio goods, and so on.
is a simple example, you go to buy a mobile phone shopping guide, see you come to know you buy a mobile phone, this is the category of goods; you said, look at the pink apple, to more memory, guide bring 128G iPhone7 red, pink is the attribute, the memory is not used to guide parameters, because Apple is fruit. Keywords iPhone. When you decide to buy pink 128G iPhone7, shopping guide and you say, today buy a mobile phone, plus 10 yuan, you can buy a mobile phone shell, this is a portfolio of recommendations. So is the electricity supplier system, step by step to guess the user thinking, users can trust website. At present, the electricity supplier in pure use of CB algorithm is not much, and for the initial establishment of the site, without user data premise, mainly rely on the CB algorithm recommended merchandise.
2, collaborative filtering algorithm (CF)
(1) user based CF
finds adjacent neighbor users based on the preferences of the user, recommending the preferences of the neighbor user to the current user. During the period of University and you often watch movies together and your bestie said, the recently released "wrestling," Dad is very good-looking, will make you want to go to see the film, because you know, her love to see, you also love to see. "People who like XX also like" is the typical User CF.
(2) object based CF
finds similar items based on the preferences of the user, and recommends similar items to him based on the user’s historical preferences. Often encountered when you buy pants, shopping guide and you say, this is our best sales, just bought one. In addition to object associations, increase >