Awesome
This repository contains the code for the paper: <br> <br> Superiority of Simplicity: A Lightweight Model for Network Device Workload Prediction <br> <br> Preprint version @ arxiv <br> <br>
Data preparation
First unpack the data/data.tar.gz archive. The contained training_series_long.csv must be located in the data directory. <br> <br>
Run the script
Python 3.6 is required to run the script.
To run the script simply do:
<br>
<br>
python code/run.py
<br>
<br>
All 10000 series will be predicted. This might take a while
(~40 hours on one Nvidia Titan GPU, will run forever on CPU).
<br>
<br>
Alternatively it is possible to predict a subset of series.
<br>
<br>
python code/run.py --start 0 --end 10
<br>
<br>
This can be used for testing or for parallelization by running
this script several times and defining respective start and
end indices.
<br>
Example:
python code/run.py --start 0 --end 2500python code/run.py --start 2500 --end 5000python code/run.py --start 5000 --end 7500python code/run.py --start 7500 --end 10000<br>
This will produce 4 submission files in data folder. <br>