This is my first time interacting with neural networks due to the needs of a personal project. The problem I’m facing is that I have a web API that validates 12-digit ids for a system I’m creating. I want to check how easy it is for a neural network to figure out the patterns applied to the creation of them and train it in a way that it can predict how “valid”(basically a value between 0 and 1) a new id that wasn’t in its dataset is. My dataset is a huge amount of ids and if they’re valid or not, in the form of the following python list(I’m working on python using the tensorflow library):
dataset = [(123456789123, 1), (234567891234, 0), ...]
I want the neural network to train using this dataset and then given a new input, output a value between 0 and 1 of how likely said input is a valid id. I want the neural network to train so that it breaks down the data into digits then segments and figures out the patterns by performing calculations, and outputs the final result from 0 to 1.
Obviously said neural network will have a large number of False predictions since it’s interacting with a web API, but I want to optimize the process as much as possible instead of relying on figuring out the patterns with my weak human eyes and randomly guessing.
Also, I want the neural network to be able to continue training as it receives feedback from that web API(basically if its prediction matches if the id is valid or not).
Keep in mind I’m a complete neural network newbie so I need as much help and corrections as possible 
Thanks in advance for your time 