**************Questions*****************************
[Q1] Using DeepAR and Wavenet from the Amazon gluon-ts package to carry out demand forecasting on provided dataset.
[Q2] Using GRU and SARIMAX to carry out demand forecasting on provided dataset, and compare their difference.
[Q3] Please understand the paper “BOHB: Robust and Efficient Hyperparameter Optimization at Scale ” and apply this technique to tune the parameters of the time series algorithms in Q1&Q2.
************* Dataset and Requirements****************
There are a series of time series files, each of which is the demand data of a product, and the data length is short. The data information and prediction examples of each item are as follows.
Please note that:
• For all above questions, please make forecasting for next 12 months started with one Period_id for all provided 1146 items. You may just need to compute the ACC = min(predicted value, truth) / max (predicted value, truth) to measure the forecasting accuracy in M0, M1, M2 and M3 months for all the items. For other predicted values of M4-M11, you may just show its predicted values. Please save the results of all the items in one file only.
• You need to take use of the additional features (order-qty1&2&3) while please note that the target prediction variable is “qty” only.
• All results should be produced by Python3 programming eventually.