How to deal with large-scale sparse data?

[DynastyTHU]

I find that there is a comment in DMatrix: 'Input data must be numpy.ndarray or pandas.DataFrame'. Does it mean that I have to provide all the data to the model, even with 99% of 0? Does xLearn support some sparse format of data, such as csr_matrix() or csc_matrix()?

[gdhzLZ]

I think you can write a function to solve this problem. And i write a function to create this form libffm data. If you need, I can seed to you.