Fraud detection

from Interview Street

At Groupon we need to take steps to detect and prevent fraudulent purchases. One form of fraud is an attempt from the same user to purchase a deal more than once using different credit card information. Given a set of orders, your task is to identify the orders that fall under this type of fraud.

An order is considered fraudulent if any of the following conditions apply:

• Two orders have the same email address and deal id, but different credit card information, regardless of street address.
• Two orders have the same Address/City/State/Zip and deal id, but different credit card information, regardless of email address.

Remember, people are tricky and are actively trying to get past this fraud detector. Your code must be able to handle the following tricks:

• Email and addresses are case insensitive: bugs@bunny.com is the same as BuGS@BuNNy.COM and 123 Sesame St. is the same as 123 SeSAME st..
• The username portion of an email address can have ignored characters. A . in an email is flat out ignored, so bugs1@bunny.com, and bugs.1@bunny.com are the same email address. A + in an email means the plus and everything after is ignored, so bugs@bunny.com and bugs+10@bunny.com are the same email address.
• Street addresses often have abbreviated words. 123 Sesame St. and 123 Sesame Street are the same address. IL and Illinois are the same state. For the purposes of not making this a typing problem, you can assume that the only abbreviated words you need to worry about are Street/St. and Road/Rd. and the only states you need to worry about our IL, CA, and NY.

We need this detection code to run quickly. The input file will be large enough so that it will behoove you to make your code as fast as possible. That said, please remember that this fraud system is part of a larger system and one that might change over time, and we expect the structure of your code to reflect that fact.

Input:

---

First line will contain a integer N denoting the number of records, followed by N lines with one record per line.

Each record contains the following information separated by commas:

• Order id (numeric)
• Deal id (numeric)
• Email address
• Street address
• City
• State
• Zip Code
• Credit Card #

Output:

---

A single line of comma separated fraudulent order ids in ascending order

Sample Input:

---

3

1,1,bugs@bunny.com,123 Sesame St.,New York,NY,10011,12345689010

2,1,elmer@fudd.com,123 Sesame St.,New York,NY,10011,10987654321

3,2,bugs@bunny.com,123 Sesame St.,New York,NY,10011,12345689010

Sample output:

---

1,2

Sample Explanations:

---

The first two orders are fraudulent, because they have the same address and deal, but different credit card information. The third order is not fraudulent because, although it shares personal information with the first order, it has the same credit card info and a different deal id.