Baseline Performance Establishment

Dear colleagues and educators,

I was just wondering when you are attempting an unprecedented classification problem

e.g. diagnosing Parkinson’s via changes in a patient’s voice
y = probability of patient developing Parkinson’s given their voice changes based on audio recordings
This isn’t possible for human physicians to do

How do you establish a baseline performance?
Is there a recommended algorithm to train and compare results against

Hi @Jamal_Hasan_Jamal_Ha,

I would look for some relevant research papers which tried Machine Learning and shared the performance and the architecture of their models. Then we can start implementing the paper’s model.


1 Like

Sometimes you just have to keep making improvements to your strategy and implementation until you get results that are “good enough” for the task you’re working on.

1 Like


Are you still working on PD? I would like to, also. Drop me a line if you want to trade thoughts.