Neural Network

Discussion in 'The Projects Forum' started by maricarbahaya, Oct 2, 2015.

  1. maricarbahaya

    Thread Starter New Member

    Aug 21, 2015
    1
    0
    Hey guys we have a project which is neural network to predict diabetes and I dont know how to do that . can you help me guys. Thanks a lot.
     
  2. GopherT

    AAC Fanatic!

    Nov 23, 2012
    6,067
    3,837
    If it were easy as fishin' I could be a musician.
     
  3. Papabravo

    Expert

    Feb 24, 2006
    10,163
    1,797
    Pick some input variables like weight, hours per week of exercise, hours per night of undisturbed sleep and build a model that gives a negative result for less weight, more sleep, and more exercise. I think it is pretty well known that more weight, less sleep, and less exercise is correlated with a high incidence of the disease
     
  4. wayneh

    Expert

    Sep 9, 2010
    12,153
    3,059
    First you need data on which to build your neural network model. If you don't have a decent set of data, you cannot really proceed. But assuming you have data, then you need neural network software to process it. Pick an application that suits your needs. There used to be several free options but that was years ago.

    How did you end up on a "circuits" forum?
     
  5. InspectorGadget

    Active Member

    Nov 5, 2010
    211
    42
    Neural networks are non-linear classifiers. They were used when the degrees of freedom on the input vector caused disjoint solution sets. I did some of the early research with neural networks in the late 80's/early 90's with handwriting and facial recognition projects.

    But they were proven in the 90s to just memorize certain training sets and could not generalize. If you have an educator using a neural network, you should let him know that in the 21st century, non-linear classifiers usually use Bayesian networks.

    https://en.wikipedia.org/wiki/Bayesian_network

    But like Wayneh says, that's a software issue and is OT on a circuits forum.
     
Loading...