IDG Contributor Network: Why today’s real-time economy needs machine learning | Tech News
“Machine learning” is not just a buzzword for futuristic applications; it is the concept of machines carrying out tasks on their own that would typically require human intelligence. Its emergence is very much happening now. It is at the top of Gartner’s hype cycle. In fact, Gartner predicts that by 2022, more than half of data and analytics services will be performed by machines instead of human beings, up from 10 percent today. And while not all machine learning use cases include real-time analytics, there is a definitive growth trend in the market for real-time decision making powered by machine learning.
Why do you think that is occurring? In my opinion, this is driven by today’s real time economy, and the fact that more than 60 percent of people increasingly feel that waiting for something that should happen instantaneously impacts their perception of the underlying brand. This is especially true if the lag affects identity or money.
Machine learning and artificial intelligence make real-time happen
Machine learning and AI are significant technologies that are emerging to revolutionize businesses, brands, and even entire industries. They have the ability to drastically reduce labor costs, generate new and unexpected insights, discover new patterns, and create predictive models from raw data. In the world of data deluge, they also have the power to operationalize data analytics and enable real-time automated decision making that wasn’t possible before. When applied to real data in an automated, low-latency manner, the results can affect business activities as they are happening in the moment. In other words, there is no longer any reason for brands to lose customers due to waiting.
A few real time decision-making use cases that have become possible thanks to machine learning are fraud prevention, personalization, and offer management for mobile games, ad bot prevention, and others. Don’t get me wrong: There are plenty of use cases that rely on machine learning to create recommendation engines—for instance, Netflix movie recommendations or Facebook offers. Machine learning coupled with real-time data analytics can have a tangible impact on high-stakes business events. Below, I explore a couple of use cases where this becomes more evident.
Use case: preventing fraud in financial services
Huawei Technologies, a communications, information, and technology solutions provider, uses a translyticaldatabase to perform real-time fraud analysis. What does this look like? In credit card transactions or mobile phone payments, every time a card is swiped or inserted or a phone is tapped or scanned, a decision about whether to authorize or decline the payment is made. This requires a recently trained machine learning model to identify fraudulent behavior based on training it received on historical fraud data. This training occurs in a big data system that receives exported information from the in-memory translytical database. The model then gets loaded as stored procedures or user-defined functions into the database several times a day.
Fraud prevention is not a one-and-done solution. Because fraudsters keep changing their methodology all the time, it is paramount to keep updating your machine learning fraud model at the same pace. This keeps the quality of decisions high and the false positive rate low. All this occurs while live credit card events are streaming with zero downtime. It’s important to note that prevention differs from detection—and allows banks to be proactive about catching fraud in-stream rather than after the fact, leading to higher customer satisfaction scores and reduced financial exposure. It’s not just a cost-saver for the financial institutions, it also helps them maintain a high brand value by minimizing exposure and preventing fraud at it occurs.
Use case: reducing fraud in ad tech
Much like banks, ad tech providers must deal with fraud quickly. Much of the damage in mobile advertising, and consequently in mobile e-commerce, is caused by ad bots—malicious code bits that behave like people to commit fraud. Ad agencies and advertisers lose millions of dollars annually, and ultimately take a hit to their brand reputation as a result of internet fraud rings like Methbots. The bots can spoof popular video content on which publishers sell ad space. They then simulate a person interacting with the video with mouse movements and fake social media information. Add to that click fraud, which occurs when a fraudster clicks (manually or automatically) on an advertisement with the intention of inflating click numbers. To detect and deal with click fraud in real time, advertisers need to monitor each click, detect anomalies, and respond appropriately.
To address these issues, the solution must be fast, accurate, and flexible enough to keep up with modern fraud attacks. Detecting and stopping this type of fraud requires a database capable of ingesting large flows of both legitimate and fraudulent traffic and deciding which traffic falls under each category—before authorizing ad spend.
WhiteOps is a provider of bot detection and human verification on the internet. It detects sophisticated bot activity and other automated threats that lead to widespread ad fraud, which costs agencies, advertisers, and brands millions of dollars annually. As part of its real-time verification program, WhiteOps processes a trillion transactions a month in its database looking for fraud activity. Detecting and preventing ad bot fraud includes spotting bots by differentiating bot action from authentic human behavior. The company’s solution provides human verification of web-scale advertising activity, providing a real-time decision for any ad impression on the internet in 10ms or less.
Combining machine learning and AI has enabled companies to detect data anomalies in 5ms to 10ms. The goal is to make a decision based on information as it comes in and even anticipate what the result will be. And herein lies the beauty of this technology union: Together, AI and machine learning are powerful tools. Add to that a fast in-memory translytical database, and the results are significant advancements within many areas of business. As the ability to predict and prevent becomes the difference between success and failure, the value of real-time surges—and promises to solve problems that have yet to be uncovered.
This article is published as part of the IDG Contributor Network. Want to Join?