Fraud detection is as simple as finding an outlier in the data given to you. For example, determining a purchase done with a credit card is out of the natural habits of the user who owns it or not. Real-time fraud detection is finding the real-time by using various fraud detection algorithms whether the purchase is legitimate or not. For example, a man buying various electronics items online through his credit card suddenly has a purchase of women’s item in a different city far away from his home. This is clearly an outlier as it does not match the purchasing habits of the man and the credit card might be blocked immediately or the man will get an immediate call for verification of the purchase.
Fraudulent transactions can cost millions per year in the financial sector. Stolen credit card and identity theft remain amongst the top list. It is necessary to solve these issues in a particular time frame. Streaming analytics comes into play as it analyzes real-time data with the help of continuous queries. It helps companies to analyze the data as soon as it becomes available to them. The company can analyze this data on a real-time basis and can find new business opportunities to generate more revenue. Streaming analytics can perform millions of queries per second. In case of our example, streaming analytics will use the historical information of the credit card user which may include the average number of transactions done per month or week, the average amount spent, types of the item purchased, locations where the items are purchased. For a particular transaction, all these queries should be calculated in a particular time frame until that particular transaction is occurring to prevent the fraudulent use of the card.
Figure 1.Streaming Analytics
Streaming analytics has various industry usages to analyze the high volume of data inflow and take action towards it. Some other applications apart from fraud detection can be real-time stock analysis, analysis of data generated from IoT devices, CRM applications, and Web Clickstream analytics.
How does Streaming Analytics work?
Figure 2.Working of Streaming Analytics
Streaming analytics starts with the production of data which may occur from one of the many devices like mobile devices, sensors, IoT, Web clickstream and transactions that is ingested in the data center. This data is that passed through the streaming analytics where the data is thoroughly analyzed with millions of queries done within a second to see patterns, a correlation in the data. This sends an output to perform an action according to the condition set in advance which could be changed as per industry.
A simple flowchart to explain the process of Stream Analytics with the example of Credit Card Fraud Detection:
The number of cases detected in cyber frauds in India was 9500, 13083 and 16468 in the year 2013-14, 2014-15 and 2015-16 respectively. As Digital India is on the rise and also the effect of demonetization has led to increase in the number of online transactions, the number is very likely to go on high and it is essential that the streaming analytics be implemented in the financial sector of India. Also, to generate more revenue and increase more business opportunities in other industries, streaming analytics can be very helpful as it will help you to gain a strategic advantage. Some of the top vendors in the streaming analytics platform are IBM Streams, Azure Stream Analytics, Data Torrent, and Oracle Stream Analytics.
-Tejas Shah (2016-18) and Aparna Gondane(2016-18)