Credit Application Process: Accept or Reject the Application

The Process

This example will walk through the process of Credit Application. This process typically involves a client filling out a credit application form, specifying details about themselves, the request loan amount, and supplying supporting documentation, such as bank statements and proof of identification.

This example process will include the following steps:

Extracting data from a Credit Application Form.
Extracting data from a corresponding Social Security card.
Extracting data from a corresponding Bank Statement.
Comparing results from step 1 with the results of step 2 and step 3, and ultimately approving or rejecting the application.

How to write Custom Service Code to compare values from the different services.

At this stage, all the data has been extracted from all the relevant documentation. The business logic to approve or reject the application should be implemented. In this example, The Social Security number written on the application form will be compared with what was extracted by the SSN service. Similarly, the salary that was written on the form will be compared with what was extracted by the Bank Statement Service.

If the Social Security numbers don't match, the application will be rejected. Similarly, if the salaries don't match within a threshold of 20 %, the application will be rejected. Otherwise, if both fields match the application will be approved.

In order to write custom code that will execute after document data has been extracted, a Custom Service Code utility must be added to the service. Please see Custom Service Code to add this utility and an overview of object data types and method prototypes.

Code sample walkthrough (C#)

This section will discuss in detail the code that is required to implement the rejection logic of the Credit Application process, as explained in the previous section.

Check the number of docs linked to the service and skip Custom Service Code execution if no documents exist.
```
if (docs == null) return new AIForged.Services.ProcessResult(docs);
```
Add some logging to indicate to the user that the processing has started by using logger.LogInformation.
```
logger.LogInformation("{stpd} Start", stpd.Name);
```
Get the Parameter Definitions of the specific fields of interest specified on the Credit Application Form, and compare the extracted values with what is provided by the supporting docs, Bank Statements, and Social Security Card. In this sample, it is required to iterate through the transaction items extracted from the Bank Statement service. To this end, the Parameter Definitions for the Description and Amount columns in the Transactions Table of the Bank Statements service. The Parameter Definition IDs of these columns can be used to retrieve the Parameter Definition objects by calling module.FindParameters.\
```
IParameterDef pdDescription = module.FindParameterDef(78090);    //Replace 78090 with your Paramdef ID
IParameterDef pdCredit = module.FindParameterDef(78092);        //Replace 78092 with your Paramdef ID
IParameterDef pdAccountHolderName = module.FindParameterDef(78148);    //Replace 78148 with your Paramdef ID
```
In order to save the values of the Parameters defined by the Parameter Definitions, a Custom Dataset is created in order to share the extracted values between services. This dataset needs to be created using the Settings View of the Parent Service of the Custom Service Code.\

Click on Create Custom Dataset in the Service Parameter view.\

Enter a name for the new Custom Dataset, for instance "BankStatementsDataSet" and click Submit.

This will create a Parameter Definition for the BankStatementsDataset Custom Dataset. Note the ID of the Parameter Definition.\
Add column names for Name and Salary to the Custom Dataset. **** Double-click on the BankStatementsDataset Custom Dataset to open the dialogue.\

Enter Name in the Column Name textbox, and click on New Column to add a column for Name. Repeat for the column name Salary.
As mentioned, this Custom Dataset was created for the Parent Service of the Custom Service Code. The Parent Service can be retrieved in code in order to access the Custom Dataset by calling module.GetParentService. The Custom Dataset object can be retrieved by calling module.GetDataSetByDef.\
```
var parentservice = module.GetParentService();
ICustomDataSet dataset = module.GetDataSetByDef(parentservice, 78817, false, false, null, null, null, null); //ID 78817: BankStatementDataset
```

To retrieve the column Parameter Definitions in a Custom Dataset, make use of modulde.FindField.

ParameterDefViewModel name = dataset.FindField(78820);  // Name
ParameterDefViewModel salary = dataset.FindField(78821);   // Salary

Iterate through all the documents that was processed in the batch. Add a try-catch clause to handle any exceptions:\

foreach (IDocument doc in docs) 
{ 
  try 
  {
    // Custom Code logic per document comes here

  }
  catch (Exception ex) 
  {
    logger.LogError(ex, "{stpd} Cannot process {docid} {docfilename}\n{error}", stpd.Name, doc.Id, doc.Filename, ex.ToString());
  }
}

Try to find possible Salary entries from the Bank Statement by using the Description column parameter of the Transactions Table by calling module.GetParameters.

List<IDocumentParameter> parDescriptions = module.GetParameters(doc, pdDescription, true, null);
parDescriptions = parDescriptions.OrderBy(d => d.Id).ToList();

Add a zero-count check and return the Custom Service Code if no bank statement data was found. Otherwise a comparison can't be done.\

if (parDescriptions.Count == 0)
{
    logger.LogWarning("{stpd} No Bank Statement entries found!", stpd.Name);
    return;
}

Iterate through each Description Parameter, get the text value extracted by the OCR, in either the Document Parameter or Verification value, and determine the Levenshtein Confidence between the parameter value, and the expected text for a salary entry on the bank statement. In this case, the text "Account Transfer In" is used to denote a salary deposit, and Levenshtein Confidence of 80 % (normalized as 0.8) is chosen.

foreach (IDocumentParameter parDescription in parDescriptions)
{
    if (parDescription?.Value == null || string.IsNullOrEmpty(parDescription?.Value))
    {
        continue;
    }

    IVerification verDescription = parDescription.Verifications.OrderBy(v => v.Id).LastOrDefault();            
    string descriptionText = parDescription?.Value ?? verDescription?.Value;

    var distance = LarcAI.Core.Utilities.LevenshteinDistance(descriptionText, "Account Transfer In");
    var conf = LarcAI.Core.Utilities.LevenshteinConfidence(descriptionText, distance, "Account Transfer In");

    if (conf >= 0.80)
    {
        // Add Logic based on the Levenshtein confidence value
    }
}

If the Levenshtein Confidence is greater than 0.8, a parameter from the Transactions Table has been matched as a salary deposit entry. The next step is to get the extracted values from the corresponding Credit column form the Transactions Table. This is accomplished by using the table Index of the matched Description parameter, and to find the corresponding value extracted from the Credit column. This is done by calling module.GetParameters and specifying the Credit Parameter Definition and the table Index.
```
IDocumentParameter parCredit = module.GetParameters(doc, pdCredit, true, parDescription.Index).OrderBy(d => d.Id).ToList().LastOrDefault();
IVerification verCredit = parCredit.Verifications.OrderBy(v => v.Id).LastOrDefault();
string creditText = parCredit?.Value ?? verCredit?.Value;
logger.LogInformation("{stpd} {var} val {val}", stpd.Name, salary.Name, creditText);
```

The Account Holder Name parameter was also extracted from the bank statement and is used to build up the record to be added to the Custom Dataset.

IDocumentParameter parName = module.GetParameters(doc, pdAccountHolderName, true, null).OrderBy(d => d.Id).ToList().LastOrDefault();
IVerification verName = parName.Verifications.OrderBy(v => v.Id).LastOrDefault();
string nameText = parName?.Value ?? verName?.Value;
logger.LogInformation("{stpd} {var} val {val}", stpd.Name, name.Name, nameText);

If either of the Account Holder Name or Credit text could not have been extracted for the current document, then the document is skipped and the next document is processed.
```
if(string.IsNullOrEmpty(creditText) || string.IsNullOrEmpty(nameText))
{
    break;
}
```
14. 15.

Code Sample (C#)

//Custom Code: Please refer to documentation
if (docs == null) return new AIForged.Services.ProcessResult(docs);
logger.LogInformation("{stpd} Start", stpd.Name);

var parentservice = module.GetParentService();

IParameterDef pdDescription = module.FindParameterDef(78090);
IParameterDef pdCredit = module.FindParameterDef(78092);
IParameterDef pdAccountHolderName = module.FindParameterDef(78148);

// Get the dataset
ICustomDataSet dataset = module.GetDataSetByDef(parentservice, 78817, false, false, null, null, null, null); //78817: BankStatementDataset
ParameterDefViewModel name = dataset.FindField(78820);  // Name
ParameterDefViewModel salary = dataset.FindField(78821);   // Salary

logger.LogInformation("{stpd} No of Docs {no}", stpd.Name, docs.Count);

foreach (IDocument doc in docs)
{
    try
    {
        // Look for possible Salary entries
        List<IDocumentParameter> parDescriptions = module.GetParameters(doc, pdDescription, true, null);
        parDescriptions = parDescriptions.OrderBy(d => d.Id).ToList();

        if (parDescriptions.Count == 0)
        {
            logger.LogWarning("{stpd} No Bank Statement entries found!", stpd.Name);
            return;
        }

        // For each description in the Transactions table
        foreach (IDocumentParameter parDescription in parDescriptions)
        {
            if (parDescription?.Value == null || string.IsNullOrEmpty(parDescription?.Value))
            {
                continue;
            }

            IVerification verDescription = parDescription.Verifications.OrderBy(v => v.Id).LastOrDefault();            
            string descriptionText = parDescription?.Value ?? verDescription?.Value;

            var distance = LarcAI.Core.Utilities.LevenshteinDistance(descriptionText, "Account Transfer In");
            var conf = LarcAI.Core.Utilities.LevenshteinConfidence(descriptionText, distance, "Account Transfer In");

            if (conf >= 0.80)
            {
                // Get the corresponding Credit Value
                IDocumentParameter parCredit = module.GetParameters(doc, pdCredit, true, parDescription.Index).OrderBy(d => d.Id).ToList().LastOrDefault();
                IVerification verCredit = parCredit.Verifications.OrderBy(v => v.Id).LastOrDefault();
                string creditText = parCredit?.Value ?? verCredit?.Value;
                logger.LogInformation("{stpd} {var} val {val}", stpd.Name, salary.Name, creditText);

                // Get Account Holder Name
                IDocumentParameter parName = module.GetParameters(doc, pdAccountHolderName, true, null).OrderBy(d => d.Id).ToList().LastOrDefault();
                IVerification verName = parName.Verifications.OrderBy(v => v.Id).LastOrDefault();
                string nameText = parName?.Value ?? verName?.Value;
                logger.LogInformation("{stpd} {var} val {val}", stpd.Name, name.Name, nameText);

                if(string.IsNullOrEmpty(creditText) || string.IsNullOrEmpty(nameText))
                {
                    break;
                }

                // Add record to dataset
                ICustomDataSetRecord dsrec = new CustomDataSetRecord();
                dsrec.SetValue(name, nameText);
                dsrec.SetValue(salary, creditText);

                await module.SaveDataSetRecord(dataset, dsrec);

                logger.LogInformation("{stpd} Saving dataset record", stpd.Name);
                module.SaveChanges();
            }   
        }
    }
    catch (Exception ex) 
    {
        logger.LogError(ex, "{stpd} Cannot process {docid} {docfilename}\n{error}", stpd.Name, doc.Id, doc.Filename, ex.ToString());
    }
}

Credit Application Process: Accept or Reject the Application

Credit Application Process: Accept or Reject the Application

The Process

How to write Custom Service Code to compare values from the different services.

Code sample walkthrough (C#)

Code Sample (C#)

results matching ""

No results matching ""