I have a very sparse database with more than 200000 rows (instances) and 500 columns that lead to almost 100 million entries. However, only 205000 of the data are non zero, that is almost 0.2% of the database.
I would like to build up a classification problem using this dataset and a label vector. I have already tried to implement a Random Forest Classifier but with poor results. This leads to my question:
What is a good machine learning model for a very sparse and highly dimensional feature set?