Log in / Register
Home arrow Business & Finance arrow Fraud and fraud detection
< Prev   CONTENTS   Next >


Applying the Z-score test on the entire payment file and indexing the Z_SCORE field by descending order, we can see from the score how much the amount deviates from the center amount shown in Figure 8.8 .

While useful, the Z-score test is more effective when applied to certain classes such as on a specific agency or specific vendor.

Z-Score Test of All Payments

FIGURE 8.8 Z-Score Test of All Payments

Let's summarize by vendor to select some to apply the Z-score test to. The new summarize file is named "Summarize by Vendor—Payment."

Since we do not have vendor numbers and are relying on vendor names, in Figure 8.9, we note that there may be issues with how the names are entered. This should be kept in mind when applying additional tests later. For instance, record 158 and 159 are likely the same vendor. Similarly, records 189 and 190 are the same vendor.

We will combine the vendors A Better Life Home Care with A Better Life Homecare and then apply Z-score to both as if they were one vendor. The new file can be called "A better life home care—combined."

You can see how the Z_SCORE1 field calculated on the combined vendor deviates from the center much more than when the Z_SCORE field was calculated against the entire payment data set in Figure 8.10. The full payment data set had very large numbers in both the negative and positive directions that made our vendor maximum of $27,496.22 seem irrelevant.


Even dollar values paid are usually of interest whether they are journal entries or actual payments by various tender types such as checks or electronic fund transfers. Most liabilities are not of even amounts and some payments include taxes, which make even amounts more unusual. We can do analysis on those that are of even hundreds, thousands, or tens of thousands. Using the payment file, we extract amounts that are in even thousands of dollars for review using the equation of (PAYMENT_AMOUNT % 1000) = 0 .AND. PAYMENT_AMOUNT < > 0

Summarizing by Vendors' Names

FIGURE 8.9 Summarizing by Vendors' Names

The % is not a percentage but instead represents Mod or modulus in the equation editor. Modulus returns the remainder when one number is divided by another. For example, where 7 mod 2, the remainder is 1 (7 divided by 2 = 3 with a remainder of 1). Where 8 mod 2 divides evenly, the remainder is zero. For our even thousand equation, if our amount divided by 1,000 has a remainder of zero, then it matches the first part of

Z-Score of Payments to a Specific Vendor

FIGURE 8.10 Z-Score of Payments to a Specific Vendor

our equation. We have the second part of the equation to exclude where amounts were zero. This eliminates excessive irrelevant records.

Even Thousand Amounts in the Payment File

FIGURE 8.11 Even Thousand Amounts in the Payment File

The result of this analysis contains 13,827 records in Figure 8.11. You can quickly scroll through some of the larger records both to get a sense of what is included and to see if there are any transactions of interest. Since we have so many transactions that meet our even thousand dollar amount criteria, we can employ sampling techniques to obtain a smaller representative sample of these records. For any general analysis that results in too many records to review in detail that cannot be further reduced by additional criteria, the sampling techniques outlined in Chapter 4 may be used.

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science