Hi I am wondering if there is any way to extract the algorithm of validity of certain codes, naming, National ID code, ZIP Code, Bank Account Number or etc.

an example would be like the following :

National Code: 5338979562
Parcel Tracking Code: 6354600321651303210321630365

the code represents a Sample Invalid National code. the length is Valid but the number, like License Plate should follow certain regulations to be accepted as valid code. and also the one i am looking for is something like:

Parcel Tracking Code

and its length is fixed:

print len(ParcelTrackingCode)

>> "25"

but not exactly this long.
my question is, in what approach in Neural Network can i find the solution to train my network in such a way that it can extract the pattern and algorithm behind the code and after training i want to check other (random/fake or real) ParcelTrackingCodes to see if they are valid or not(prediction)? to be more specific what type of Neural Network should i use ?


my requirements is to be able to validate the Parcel Tracking Code and if it is valid i want to allow user to perform certain tasks but if it is not valid i prevent them advancing next steps.

note: i have around 20,000,000 records containing Parcel Tracking Code

CodePudding user response：

Since you have a big dataset, you can try the Generative Adversarial Networks (GAN). But the problem is that the resulting network will not be 100 percent accurate. So you can not rely on it.

Also, while some ID codes are completely generated by a formula, others are partially or totally generated from random numbers or the current timestamp. So basically, they generate a code and store it in a database. And during the validation phase, they actually look up the database for the code. This way, they make it very difficult to generate valid codes.