The code and data are provided in this gist : https://gist.github.com/987394
There is a significantly large variation in classifier accuracies while using sparse.LinearSVC and sparse.SVC despite perf…
I'm not sure it is normal or not. Just I hope to check.
svmfitted=LinearSVC(loss='l2', penalty="l1", C=1000,dual=False, tol=1e-3).fit(X_train, y_train)
pred = svmfitted.predict(X_test)
print 1.0*sum…