In the our research article, we investigated the performance of machine learning approaches on XRD datasets. The machine learning models had to classify crystal systems of materials. Instead of utilizing the tabular data format, the study employed two types of datasets[rendered images]: graphical and pixelated image datasets. The images datasets performed very well as compared to the XRD tabular format dataset used by other researchers in their previous studies. By training on the images dataset(graphical and pixelated images) our machine learning models were able to classify the crystal system of material with 98% accuracy while testing them on an unseen dataset.
This technique can be used in any large data set and implemented on CNN to increase the accuracy compared to statistical models.
Because CNN and other neural networks work well on images rather than CSV file base data, so it’s a good approach to go with, for better model accuracy.
Tablular data single rwo would be look like this after transforming into images and graphs.
