Mapping theory to the real world is not always the most intuitive process imaginable. There are several successful data mining applications that have been deployed across various sectors. In this section you learn examples of real-world applications which use data mining technology.
Have you ever received a call from your credit card company asking whether you made a specific credit card purchase? Do you know why you received the call? Chances are very good that it was due to an anomaly detected in your credit card usage as part of the company's fraud detection effort. Typically, customer usage patterns on credit cards are quite consistent. When a credit card is stolen, the usage pattern changes drastically. In spite of increasingly advanced theft protection schemes, credit card companies still lose a lot of money due to theft. Because credit card fraud is roughly 10% higher on the internet than off, Visa introduced CISP (Cardholder Information Security Processing) in 2000 and MasterCard followed with its Site Data Protection Service (SDPS) in 2001. The CISP and SDPS only help in securing and validating the data and do not actually prevent the use of stolen credit cards. In order to detect anomalies and act immediately, credit card companies are now using data mining to detect unusual usage patterns of credit cards; and once such a pattern is detected, the customer is called to verify the legitimacy of certain purchases.
Now here is an example almost everyone can relate to. Have you shopped at Amazon (the online book seller) and seen a suggestion pop up that read something like this: "Customers who bought this book also bought the following" and then some list of pertinent books followed? Do you know how they do this? This is typically accomplished with the use of a data mining algorithm called "association rules." In order to boost sales, companies like Amazon use this algorithm to analyze the sales information of many customers. Based on your book buying behavior, Amazon uses the algorithm to predict what other books you would likely be interested in. From the list of books provided by the algorithm, they typically choose the top 5 books that have the highest likelihood of being purchased by the customer — then they suggest those books. Another example of where just such an algorithm is being used is in the area of DVD rentals.
As many of you sports fans know, NBA coaches need to analyze opponent teams and adopt appropriate strategies for winning future games. Typically, the coach will look for key players on the opposing team and appropriately match up his own players to counter their strengths and expose their weaknesses. Relevant information can be from past games that have been analyzed and gleaned from other sources. The NBA is fast paced, and coaches need to adapt based on current game situations. For this purpose, they need to analyze information every quarter and often in real time.
NBA coaching staffs collect all the information on the players and points scored during a game and feed it into a data mining software application called Advanced Scout. With the help of this software, coaches are able to analyze patterns — when did the opponent score the most points, who were the players on our team, who was guarding the highest point scorer on the opposing team, where were the shots taken, and so on. With such information readily available, coaches adapt to the situation and make decisions that will help their team to win.
Yes, but how was Advanced Scout helpful, you ask? When the Orlando Magic NBA team was devastated in the first two games of the 1997 season finals, which was against the second-seed Miami Heat, the team's fans began to hang their heads in shame. Advanced Scout showed the Orlando Magic coaches something that none of them had previously recognized. When Brian Shaw and Darrell Armstrong were in the game, something was sparked within their teammate Penny Hardaway — the Magic's leading scorer at that time. Armstrong was provided more play-time and hence Hardaway was far more effective. The Orlando Magic went on to win the next two games and nearly caused the upset of the year. Fans everywhere rallied around the team and naysayers quickly replaced their doubts with season-ticket purchases for the following year.
Companies spend a lot of money on call center operations to meet customer needs. Customers use the toll-free number provided by the company and the company pays for each call based on the duration of the call. Typically, most calls target a few specific questions. For example, if the documentation for product setup was not sufficient, the call center might get calls with the same question or related questions on getting the product set up and configured properly.
Often the information obtained from customers is entered in the computer system for further analysis. With the help of Text Mining, the customers' questions can be analyzed and categorized. Most often you would end up identifying a set of questions that are due to a specific problem. Companies can use this information to create a FAQ site where they can post answers on how to solve the specific problem. Making the FAQ available and providing the answers to some of the common problems helps the company and the customers have a faster turnaround, saving both time and cost. In addition to this, the call center operators can be trained to use the information provided by Text Mining to easily nail down a solution to the problem posed by the customer. The duration of each call is reduced, thereby saving valuable cash for the company.