Table 2. MAPE for 1D-CNNs for each crystal system.

Null prediction refers to a MAPE prediction based on the mean lattice parameters of the data set. ML full corresponds to the prediction from the 1D-CNN models for 0–90° in 2θ. ML reduced corresponds to the prediction from the 1D-CNN models for 0–30° in 2θ.

Symmetry	Null prediction	ML prediction full	Ratio 1	ML prediction reduced	Ratio 2	Training data set size
Cubic	51.49	7.55	6.8	4.08	12.6	30 705
Hexagonal	47.37	7.35	6.4	8.94	5.3	17 842
Trigonal	46.58	15.72	3.0	15.30	3.0	25 784
Tetragonal	48.77	11.56	4.2	9.09	5.4	37 183
Orthorhombic	29.94	10.06	3.0	12.92	2.3	161 087
Monoclinic	24.76	11.79	2.1	11.23	2.2	445 708
Triclinic	20.06	3.11	6.5	2.68	7.5	243 651