I have got 923 classification datasets. For these data sets, steps should be implemented below.
Classification Meta Features:
1) To produce a software which extracts classification of datasets' meta features.
2) To predict that which algorithm is the most successful on a given set of data.
3) To find generalized rules between the most successful algorithm and the meta features.
For this task, a software which saves meta features in arff format and combines with the result arff.s should be produced.
For meta features you can examine the Statlog Project and Ho's article.
Max budget: $250
Should be finished in 3 days.