Feature extraction

Hello,

I have my own data which is composed of XYZ values and the atom types. I would like to load and use it in the matminer library, but I do not know how should I format this data for feature extraction task. There is some code that I can use as an example?

Thanks.

Hi OvTuto,

You should consult the pymatgen Structure and Lattice classes documentation (http://pymatgen.org/_modules/pymatgen/core/structure.html) to get your XYZ + atom type values into pymatgen Structure objects. Exactly how you do this depends on how your data is right now, but I would guess you should either (a) first make a Lattice object based on your XYZ values and then use that as input to a Structure, or (b) just use the Structure.from_sites method to make your structures (you can input the coordinates directly I believe).

Once you have all the crystal structures as Structure objects, you can put them in a dataframe and use them as input for feature extraction. For example of how to do this, please see the introduction example on the examples repo (https://github.com/hackingmaterials/matminer_examples/blob/master/matminer_examples/machine_learning-nb/bulk_modulus.ipynb)

Thanks,

Alex

···

On Friday, March 22, 2019 at 1:13:59 PM UTC-7, OvTuto Riais wrote:

Hello,

I have my own data which is composed of XYZ values and the atom types. I would like to load and use it in the matminer library, but I do not know how should I format this data for feature extraction task. There is some code that I can use as an example?

Thanks.

See also:

http://matgenb.materialsvirtuallab.org/2013/01/01/Basic-functionality.html

···

On Sunday, March 24, 2019 at 6:05:34 PM UTC-7, [email protected] wrote:

Hi OvTuto,

You should consult the pymatgen Structure and Lattice classes documentation (http://pymatgen.org/_modules/pymatgen/core/structure.html) to get your XYZ + atom type values into pymatgen Structure objects. Exactly how you do this depends on how your data is right now, but I would guess you should either (a) first make a Lattice object based on your XYZ values and then use that as input to a Structure, or (b) just use the Structure.from_sites method to make your structures (you can input the coordinates directly I believe).

Once you have all the crystal structures as Structure objects, you can put them in a dataframe and use them as input for feature extraction. For example of how to do this, please see the introduction example on the examples repo (https://github.com/hackingmaterials/matminer_examples/blob/master/matminer_examples/machine_learning-nb/bulk_modulus.ipynb)

Thanks,

Alex

On Friday, March 22, 2019 at 1:13:59 PM UTC-7, OvTuto Riais wrote:

Hello,

I have my own data which is composed of XYZ values and the atom types. I would like to load and use it in the matminer library, but I do not know how should I format this data for feature extraction task. There is some code that I can use as an example?

Thanks.