Data Input for pattern recognition


I am currently working on an activity classification task. Currently, I have a neural network that is performing ok, but I would like to see if XGBoost can perform better.

My issue is with the data input. For training in my neural network, I have multiple csv files and each csv has a label. The model takes the csv and classifies a label. In the case of production, a csv is the input and based on that, I get a label.

XGBoost from my knowledge only takes a 2D array of data, so it seems that I cannot take my data and create a 3D array for prediction. Has anyone used XGBoost for a pattern recognition type application with similar type data? Is this possible?

No, XGBoost only takes 2D array input. Maybe you can flatten each CSV into a single row.