"may we also suggest"

[isaac_cm]

Hi to all,
When visiting site on the net and enter into product detail page that contain a full description of the product and a big picture about it, I see also a commercial says "may we also suggest" offering to buy also two or three more products related to the current one

and I did not know the logic way to do that, I have all my products belong to main/sub categories so I keep thinking if I have to create two more fields to write down manually the code of the other two related products to the one I browse,

or do I have to define only a related main/sub category to the current product and the related products will be displayed randomly from the main/sub category I choose, or there is another way I have to use !!!?

(I hope it was clear)

please advice
thanks

[isaac_cm]

BUMP

[Eric_Ryk]

Well you could either log what people buy in groups or sessions and try to map it. Or with each product you could include keywords and find other products with the same keywords.

[isaac_cm]

by "keyword" or by "category", I saw amazon display an offer of two books related to the one I choose !!!

there is a lot of my products that did not have a common name
Are you sure it is based on product name keyword not category ?


thanks

[keeB]

isaac, obviously it depends on what you sell.

Amazon probably uses a few different approaches. They track who searches for what, and track everywhere you go. So if a user goes to buy book A then goes to book B, they may suggest that one to the next user.

They may also group by Author, Genre or Category, Type of product (electronic) component (like a coffee mug, then they might suggest some coffee to go with it) etc, etc, etc.

[Eric_Ryk]

Keywords are not the names of products. Lets say you have foozball and air hockey. Neither of those share any title similarities, but they would both have the keyword table-top game. So you personally make a list of keywords for a product when adding it to the list of products.

[isaac_cm]

[quote]Keywords are not the names of products. Lets say you have foozball and air hockey. Neither of those share any title similarities, but they would both have the keyword table-top game. So you personally make a list of keywords for a product when adding it to the list of products[/quote]

I see now what you mean, but now I need advice in how to do it (theoretically) , I need only to suggest two more different products to the current selected one, should I create two more fields in "product admin page" to include them manually  or make the hole thing work randomly ?

by the way the site if for gifts (footwear, bags, jewelery, etc... )

thanks to all

[Eric_Ryk]

Well what you can do is just create a field for keywords, and then find the 2 results with the highest number of matching keywords.

[isaac_cm]

many thanks

[isaac_cm]

I am gona need some help on the sql to do that
http://www.phpfreaks.com/forums/index.php/topic,117981.0.html

if there is a tutorial do what I want it will be appreciated

thanks

[utexas_pjm]

Another thing you may want to look into is association rule anlysis.  I assume you keep track of what users are buying from your site.  You can use this data to construct pretty powerful association rules that will in essence give you a "people who buy this also buy this" type of relationship.  This will only work once you have a fair amount of transactional data to mine and might be a little bit more work than you wanted to invest but I thought I'd throw it in anyway.

http://dms.irb.hr/tutorial/tut_assoc_rules.php

[isaac_cm]

many thanks utexas_pjm but as you said I need only a product recommendation system for similar products , if you a tutorial do that I will very appreciated

thanks

[utexas_pjm]

If you're looking to suggest products based on similarity then the biggest hurdle for you will be to define how you want to measure similararity.  That is, how do you determine when product A is more like product B than product C.  Once you've defined such a measure you can build a simple K-nearest neighbor algorithm based on cosine similarty or euclidean distance.  As the name implies the algorithm will return to you the closest K items to the chosen item. 

[isaac_cm]

but how I can build this algorithms I need some sort of tutorial on that issue