I am developing a System to alleviate errors in invoice generation system. Data for generating Invoices - may not be available many times, which causes errors. How are Deep Learning models used on structured data (data in databases) ? What models can I use for this ?
Hi, @neduljee, and welcome to the community!
Please describe your problem in more detail, and, if it’s OK, share a few data samples with us.
Hi Yurij Mikhalevich,
So the problem we are trying to solve is that of generation of bulk invoices that relay on data to be gathered form various datasources. There is no explicit template for an invoice as it depends on client type, geos that the client belongs to, country that can have different rules and laws governing business, the service that the client has used and how much of it, the currency that they need to be billed in. Mainly as we near the deadline for generating these invoices all the data is in a database with various tables. I am trying to develop a service that can pinpoint any data that is missing that will create a failure to generate these invoices. So I am trying to model this as an anomaly detection paradygm where the successful invoices are the ground truth. Would this be a correct way to proceed ?
Thank you for providing more details about your problem.
If I understand it correctly, you have to generate various invoices for different types of clients operating in different countries, and these variables affect whether you can generate an invoice or not. And you need to build a system that will alert about missing data for a particular type<>country combination. Is this correct?
Why do you want to use ML for that? If you have a fixed number of types of clients, and you know beforehand which kind of data you need to generate an invoice for a client with a given type<>country combination, why not just code the algorithm that will check for missing data?
It is true that there are a fixed number of clients - but this number is very large and the number of invoices to be dealt with is also very large - more importantly there is no fixed template for an invoice - it can differ for product, service, country, geography etc. So we want a system which can quickly identify the missing pieces from a previous similar invoice that was correct. While the pieces of info are in a database - there is no fixed structure to an invoice.