Demand forecast using time series data

[anwar saidan]

Hello, 

I am working on demand forecast in supply chain. I have a dataset of weekly product quantity spanning 3 years, resulting in 158 observations for each of my 1000 products.

So, my dataset has a dimention 158000*4 because I have 3 other features, other than the 'quantity' feature that I want to predict.

I want to create a prediction model that predicts the 'quantity' in the future given then caracteristics of the product, that figure in the 3 features in my trainning set. 

Can I apply ARIMA to my hole 158000*3 dataset, or is it unpossible because each date figures 1000  times in my dataset ? 

Or shall I train the model for each product separatly ? 

Thank you for your support