Keras/Tensorflow, supervised - regression, deep learning neural network with GUI. Can you do this with limited datasets?

Cancelled Posted 5 years ago Paid on delivery
Cancelled Paid on delivery

See attachments for short explanation however I really recommend reading this below before contacting me.

1. The ANN should learn from the data parameters to predict the numerical order of the INDEX labels that will happen in a future event. It should also give an estimation percentage of prediction accuracy. For initial training the known correct answers are given for learning but this is NOT the main method of operation. Normally the the ANN will make a prediction and then at some point in the future the correct answers might be given. It's important to understand that prediction and learning are two operations which are performed at separate times.

As far as I can determine this is a supervised - regression, deep learning neural network scenario but if you have some other idea to do this better let me know. I would like to have as many hidden layers as my pc can still function with in seconds or minutes, not hours. I have a skylake 8400cpu, nvidea 660Ti cpu. Whether using the gpu with cuda or just the cpu, I don't know which is better.

2. The GUI should be something like shown in the attached picture. I will also need your help explaining setting up the environment for the program, if I need cuda etc...

3. Please don't ask what this project is specifically about. Having said that it might be important to know for neuron weighting that these predictions will always be approximate because the correct answers will have a large amount of variable outcomes. Example: Like a marathon with multiple runners, we could guess the fastest person should win but someone might not feel like being there and so doesn't try hard, maybe falls down or whatever.

I have filled in all missing historical data with generic numbers like "0", "50" or "1000". This isn't ideal but I can't seem to get a straight answer of how to deal with missing data so I added some numbers which at least will be consistent. At the moment I have about 100 data sets with about 1800 rows, but working on more. The three attached are examples to show the layout. I will continue to train the ANN after your initial setup.

5. PAYMENT: This is a full payment when I see the program working on my pc. Please don't ask for a milestone part payment or something else. I have always paid for the work done on freelancer projects before so if you don't trust I will pay you then with respect don't bid.

Thank you for reading this,

Rob

Algorithm Artificial Intelligence Machine Learning (ML) Neural Networks Python

Project ID: #17499897

About the project

3 proposals Remote project Active 5 years ago

Awarded to:

dilichan

Hey Rob, I think we have spoken to each other before. I understand your problem up to some extent. Hopefully we can discuss more about this over chat. Regards, Dilina.

$150 USD in 10 days
(7 Reviews)
4.0