Recommender System without collaborative filtering

I wrote a technical article "Recommender System without collaborative filtering" in English (not in German) with Python (not R). This article is an extended version of the previous article.

The aim is not changed, but there are lots of improvements.

  • The source codes and data sets are provided at my GitHub page.
  • Porter's stemmer and a list of stop words are used.
  • No clustering analysis (because of the result of the previous article) is applied.
  • Comparison of different predictive models with different parameters is given.
  • Correlation matrix is calculated.

If you have any comment, please leave a comment here or send a message on twitter (@stdiff)!

[edit] I did not give a random seed for a few (not all) estimators of random forest. So the result which you get on your machine could be slightly different from my result.

Share this page on        
Categories: #data-mining  #development