I have written an article about TFIDF and machine learning in German. This is an examination for implementation of my recommender system (without collaborative filter).

• Training Data: web pages with labels (±1)
• Test Data: web pages without labels.

I have tried several method: SVM, random forest, penalised logistic regression, k-means, hierarchical clustering, cosine similarity and ratio of distances to labelled data.

If you have any question, suggestion for machine learning, German, etc., please leave a comment at this entry or send me a message on twitter!