About the weight value of the associated recommendation function

In fact, many people now have http://www.aliyun.com/zixun/aggregation/8699.html "> association recommendation function, some of their own development, some with a third party (such as Mailu), embodied in two aspects:" The user who saw this product also saw what product "and" bought this product the user also bought what ". But hardly see which station makes the relevance recommendation of the accuracy of particularly good, to the network examples, related recommendations:

"Read and read" is the use of wheat road (thanks to Mai Road, thank Liang Classmate), the recommendation is basically the same price or the same brand of goods, we can think it is a regular reasonable, but think about it, a recommendation of 6 associations, users will be more dazed, do not know which of the 6 will be more in line with his habits, It's too much trouble to see all 6.

"Bought and bought" is done by himself, but very TMD is not accurate, because there is no weight, such as a memory card, the recommendation actually has cameras, lenses, tripod and the like, because they have appeared in the same order. If you look at a camera, recommend a card, this is no problem, but conversely, I think no one saw a memory card, want to buy another camera it ...

The solution is what, associated weights, Mars classmate method is preferred to recommend the high weight of goods (http://www.paidai.com/displaythread.php?boardid=11&topicid=3225), But this method has a little problem, is more black-screen, not convincing (at least on the surface, the customer does not understand why you recommend this to him). So let's move on to a different approach and recommend the associated product, give them a different "weight ratio label", such as read or bought "veryls e-commerce those things" users have 60% also read or bought the "old criminal talk about E-commerce"; 30% also saw or bought the "E-commerce Angry youth Brother" 10% also saw or purchased the "postmodern Life of the electric trader's aunt" ... In this way, we show all the relevance to the user, clearly tell the user the weight value, tell the user "public opinion", by himself to make the right choice, I think this technology should not be difficult to achieve. (Joyo is doing this, mmm.)