1
$\begingroup$

I need to predict the performance (CPI cycles-per-instruction) of 90 machines for the next hour (or day). Each machine has a thousand records (e.g. CPU and memory usage).

Currently, I am using a neural network with one hidden layer for this task. I have 9 inputs (features), 23 neurons in the hidden layer, and one output. I am using the Levenberg-Marquardt algorithm. Examples of the inputs (or features) are the CPU and memory capacity and usage, and the machine_id. and Output is performance. I have 90 machines. Currently, I get an MSE of $0.1$ and an R of $0.80$.

My dataset consist of 30 days. I trained my network for the first 29 days, and I use day 30 to test.

I have been advised to use deep learning to have more flexibility and improve the MSE and R results. Could deep learning be helpful in this case? If yes, which deep learning model could I use to improve the results?

Improve this question
$\endgroup$
2
  • 1
    $\begingroup$ No. Deep Learning efficient for big networks and samples of extremely high dimensions. It don't make sense to use it for 9 element vectors as samples. $\endgroup$
    – mirror2image
    Commented Dec 12, 2019 at 7:10
  • $\begingroup$ It's highly plausible that you don't need anything neural. For model selection, the following image is always a good rule of thumb. https://i.sstatic.net/Y8jAg.png. $\endgroup$
    – MatthewTuby
    Commented Dec 12, 2019 at 7:30

0

Reset to default

You must log in to answer this question.

Browse other questions tagged .