Comments on: Collaborative filtering made easy http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/ Bryan O'Sullivan's blog Wed, 08 Feb 2012 06:41:38 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: Matt http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/comment-page-1/#comment-328841 Matt Wed, 25 Jan 2012 17:26:15 +0000 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/#comment-328841 Thanks for the code Bryan. A warning to anyone using it, as explained in a comment in the Haskell port, the `update` method is wrong from the second time you use it. http://www.serpentine.com/blog/2007/08/27/weighted-slope-one-in-haskell-collaborative-filtering-in-29-lines-of-code/#comment-75536 Thanks for the code Bryan. A warning to anyone using it, as explained in a comment in the Haskell port, the `update` method is wrong from the second time you use it.

http://www.serpentine.com/blog/2007/08/27/weighted-slope-one-in-haskell-collaborative-filtering-in-29-lines-of-code/#comment-75536

]]> By: Matt http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/comment-page-1/#comment-328837 Matt Wed, 25 Jan 2012 15:31:34 +0000 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/#comment-328837 Hey! Your link to `slope_one.py` is dead :( Hey! Your link to `slope_one.py` is dead :(

]]> By: ç”±google reader的推荐订阅推荐看网站在线推荐算法 « 404 NOT FOUND http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/comment-page-1/#comment-310683 ç”±google reader的推荐订阅推荐看网站在线推荐算法 « 404 NOT FOUND Fri, 15 Jul 2011 08:54:11 +0000 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/#comment-310683 [...] http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/ Java： http://taste.sourceforge.net/ [...] [...] http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/ Java： http://taste.sourceforge.net/ [...]

]]> By: Slope one:简单高效的推荐算法 | web开发 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/comment-page-1/#comment-307556 Slope one:简单高效的推荐算法 | web开发 Mon, 27 Jun 2011 06:31:06 +0000 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/#comment-307556 [...] http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/ [...] [...] http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/ [...]

]]> By: anonymous http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/comment-page-1/#comment-273281 anonymous Thu, 21 Oct 2010 10:27:12 +0000 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/#comment-273281 (forget about the post above) (forget about the post above)

]]> By: anonymous http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/comment-page-1/#comment-273280 anonymous Thu, 21 Oct 2010 10:18:36 +0000 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/#comment-273280 Hi, shouldn't the line freq = self.freqs[diffitem][item] actually be freq += self.freqs[diffitem][item] ? The problems mentioned by PB are also not fixed yet ;-) Hi, shouldn’t the line
freq = self.freqs[diffitem][item]
actually be
freq += self.freqs[diffitem][item]

?
The problems mentioned by PB are also not fixed yet ;-)

]]> By: Nan Zheng http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/comment-page-1/#comment-258930 Nan Zheng Wed, 28 Apr 2010 14:09:50 +0000 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/#comment-258930 Althogh I am not familiar with Python, I still feel this blog great as it make an easy way to applicate slope one. However, I wonder if you considered the input data scale. As if the input rating data is quite huge, can the dic still works? Althogh I am not familiar with Python, I still feel this blog great as it make an easy way to applicate slope one. However, I wonder if you considered the input data scale. As if the input rating data is quite huge, can the dic still works?

]]> By: Arayana Bilgi http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/comment-page-1/#comment-258701 Arayana Bilgi Sat, 24 Apr 2010 05:57:32 +0000 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/#comment-258701 Great code here; could implement at the site flawlessly. Great code here; could implement at the site flawlessly.

]]> By: 推荐系统：Slope One 算法 » Beyond Search - 最好走的路越走越难，最难走的路越走越容易 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/comment-page-1/#comment-254151 推荐系统：Slope One 算法 » Beyond Search - 最好走的路越走越难，最难走的路越走越容易 Sat, 30 Jan 2010 20:12:56 +0000 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/#comment-254151 [...] Python 的朋友可以看这篇 Blog，“tutorial about how to implement Slope One in Python”，非常详细的介绍了 Slope One 算法在 Python [...] [...] Python 的朋友可以看这篇 Blog，“tutorial about how to implement Slope One in Python”，非常详细的介绍了 Slope One 算法在 Python [...]

]]> By: Jason DeFontes http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/comment-page-1/#comment-190049 Jason DeFontes Fri, 31 Oct 2008 17:28:45 +0000 http://www.serpentine.com/blog/2006/12/12/collaborative-filtering-made-easy/#comment-190049 @Doug, Yes, that is expected. If you put the algorithm into words (in the simplest case of only two items) it would be something like: "On average, everyone else rates Octopus with two more points than they rate Squid. Therefore, your predicted rating for Octopus should be two points more than your actual rating for Squid." If your actual rating for Squid is 4, then your predicted rating for Octopus will be calculated as 6. If the acceptable range of ratings is 1 to 5, then you just have to truncate the prediction to make it fit. @Doug,

Yes, that is expected. If you put the algorithm into words (in the simplest case of only two items) it would be something like:

“On average, everyone else rates Octopus with two more points than they rate Squid. Therefore, your predicted rating for Octopus should be two points more than your actual rating for Squid.”

If your actual rating for Squid is 4, then your predicted rating for Octopus will be calculated as 6. If the acceptable range of ratings is 1 to 5, then you just have to truncate the prediction to make it fit.

]]>