11 Cool Ways to Use Machine Learning
Machine learning is becoming widespread, and organizations are using it in a variety of ways, including improving cybersecurity, enhancing recommendation engines, and optimizing self-driving cars. Here's a look at 11 interesting use cases for this technology.
![](https://eu-images.contentstack.com/v3/assets/blt69509c9116440be8/bltfd2ea248f51a03fb/64cb461eef4b20e629075d64/board-862112_1280.jpg?width=700&auto=webp&quality=80&disable=upscale)
For years, machine learning has been used for image, video, and text recognition, as well as serving as the power behind recommendation engines. Today, it's being used to fortify cybersecurity, ensure public safety, and improve medical outcomes. It can also help improve customer service and make automobiles safer.
"Machine learning allows you to look at volumes of data and do volumes of calculations that a person really can't do," said Lisa Dolev, founder and CEO of operational intelligence solutions provider Qylur, in an interview.
Machine learning can identify patterns that humans tend to overlook or may be unable to find as fast in vast amounts of data. Organizations are using machine learning to make new discoveries, as well as to identify and remediate issues faster.
[ In IoT, there are coding languages and then there are coding languages. Which ones should you learn? Read 11 IoT Programming Languages Worth Knowing. ]
Meanwhile, deep learning, a sophisticated branch of machine learning, is gaining popularity. It requires massive amounts of data and massive processing power. In a lot of cases, deep learning is being used in tandem with machine learning to improve outcomes, such as lowering the number of false positivies in security breach detection software. One reason some organizations are using deep learning is to automate more of the machine learning lifecycle. Comparatively, machine intelligence and human intelligence are often paired to overcome the limitations of rules-based systems.
"Deep learning can detect more granular or more sophisticated [information]," said Dr. Hui Wang, senior director of global risk sciences at PayPal, in an interview. "We use it as part of our tools [because] it enables a higher degree of fraud and money-laundering detection."
In short, there are different ways of applying machine intelligence to problem-solving. Here are 11 options for you to consider. Once you've reveiwed these, tell us what you think in the comments section below. Is machine learning something your organization is looking to implement? Is deep learning on your radar? What potential use cases do you see for these technologies?
In 2014, Kaspersky Lab reported it was detecting 325,000 new malicious files every day. At that rate, humans and even signature-based security solutions can't keep up, which is why machine learning and deep learning are necessary.
"Nearly all new malware differs less than 2% from previous malware," said Eli David, CTO of institutional intelligence company Deep Instinct, in an interview. "Our deep learning model has no problem dealing with the 2% - 10% mutations we see every day."
Deep Instinct uses a large core of several million malicious files, tens of millions of legitimate files and malware that Deep Instinct may have mutated by 20% - 50% for training purposes. The more radical malware mutations make the training more difficult, but they also make the model more resilient. Once the training is finished and the synapses have been updated, a text file of the synapses can run deep learning in prediction mode.
Legal documents are often too complicated for the average person to easily comprehend. Some hire a lawyer. Others may skim the documents, or even ignore a document's content, hoping that somehow everything will work out. Some are overconfident about their ability to understand the documents. Deep learning can help.
"We built a legal language model that allows us to translate legal language into a big string of numbers. We're using deep learning and topological data analysis, which is sort of a dark corner of geometry," said Dan Rubins, founder and CEO of Legal Robot, in an interview. "A lot of techniques have been developed for natural language processing on normal language, but there's nothing natural about legal language."
In addition to translating legal language into plain language, Legal Robot can determine what's missing from a contract and whether there are elements in a contract that shouldn't be there, such as a royalty fee section in a non-disclosure agreement. When it comes to contracts, there is often a disparity of bargaining power between the parties, such as between a payday lender and lendee, or between a corporation and an individual. Because contracts usually contain a lot of complicated language, people often agree to terms and conditions that favor the other party.
PayPal is using deep learning to prevent fraud and money laundering at granular levels. By combining deep learning with machine learning and other tools, the company can precisely discern between legitimate and fraudulent buyers and sellers. According to Hui Wang, PayPal's senior director of global risk sciences, it's all about anomaly detection.
"A sophisticated algorithm can identify a peer group at a more accurate level [than] everybody selling electronic appliances. [When it comes to money laundering] people pay attention to high dollar amounts, but some organizations, such as Girl Scouts of America, are legitimate," said Wang. "Others may not be processing as much but show up as abnormal."
PayPal also uses human detectives who are constantly analyzing patterns in the system. When someone detects an anomaly, she makes up "a good story and a bad story" that explains two possible reasons for the transaction. Machine learning helps scale that up.
Improving your position in the Tour de France is difficult if you have little or no perspective into the positions and status of other cyclists. About 200 cyclists participate in the race, but not all of the riders are covered on TV. That means neither the public, the cycling teams, or the athletes lack visibility into the entire event.
"If you're not leading the race, the cameras generally aren't on you. The coaches need to know what's happening in the event beyond what's on TV, so they can give information to the athletes," said Robby Ketchell, chief data scientist and co-founder of software and device maker winningAlgorithms, in an interview. "We use what people are saying on social media to help us understand what's happening in the event."
Social media data is less reliable than some other kinds of data, however, because humans contribute both fact and fiction. The algorithm is able to determine the credibility of individuals reporting race details over social media. In the 2012 Tour de Italia, winningAlgorithms was able to tell athletes what was happening 5 minutes before the same information was broadcast over the race's radio system. It also helped a Tour de France team win one stage of the 2013 and 2014 races.
The IBM Institute for Business Value recently surveyed 175 auto industry executives in 21 countries. Seventy-four percent expect that by 2025, vehicles will self-optimize and provide advice in context. Specifically, they'll be able to learn about themselves, the surrounding environment, and the behaviors of the drivers and the occupants.
"Vehicles are intelligent today, they're becoming a little bit more intuitive, but what we see happening over the next 10 years is this notion that they'll be able to do things by themselves," said Ben Stanley, global automotive lead for the IBM Institute for Business Value, in an interview. "They'll be able to integrate themselves into the Internet of Things, they'll be able to configure themselves based on the occupants and the environments, fix themselves, drive by themselves, and they'll be able to communicate with other vehicles and networks."
Future vehicles will also be able to personalize driving experiences by observing and mimicking their human drivers/owners. The vehicle will learn by observing the driver's actions and conversing with the driver. If the driver opts in, the car will be able to share information with the automotive manufacturer and others in the industry, beyond the kinds of data that are being shared today.
Large retailers employ analysts to help identify, reduce and prevent fraudulent transactions. Many have used rules to block transactions from suspicious locations, such as Nigeria and Ukraine, but that approach is so blunt it also blocks legitimate transactions. Machine learning helps retailers and others manage fraud in a more precise way.
"The way you look at patterns is to take a lot of small data points which, by themselves, wouldn't be fraud conclusive, but when you add them up it can give you a strong [indication of fraud]," said Rurik Bradbury, CMO of ecommerce fraud prevention solution provider Trustev, in an interview. "If you imagine a website where someone comes, browses around, and puts something in their shopping cart, it would be too complex for a human to make sense of the millions of little bits of data.
The goal is to identify fraud patterns before a product ships, without delaying the delivery of products. According to the 2015 CyberSource North American Fraud Benchmark Report, 27% of the online orders a merchant receives will be routed to fraud analysts for additional manual review. Of those, roughly 85% are ultimately deemed valid and accepted.
Airline passengers, concert attendees, and sports fans have something in common: they're screened by security guards and systems. Human screeners often overlook items that machine learning can identify. And, machine learning can easily adapt to seasonal changes affecting bag types and bag contents, or the specific requirements of a particular venue.
"Venues use security to keep out bombs, guns and other items. They don't like selfie sticks because you can harm someone with it. FIFA doesn't want country flags coming in," said Lisa Dolev, CEO of operational intelligence solutions provider Qylur. "When you have 100,000 people trying to get into a stadium, you can have hundreds or thousands of false alarms. We're getting a lower false alarm rate."
Machine learning can improve the efficiency of customer service by understanding customers and their issues at a granular level. Whether the support is virtual, provided by humans or a combination of both, customers and their issues tend to be divided into very large buckets, such as Product A and Product B.
"Customer effort is directly correlated with lifetime value, so the amount of time it takes a customer to use your product or resolve a problem has a clear relationship with the amount of money they spend," said Patrick Rice, founder and CEO of predictive analytics platform provider Lumidatum, in an interview.
Machine learning can easily discern between the customers that are beginning to use a product versus those that have more experience with the product, which enables efficient customer support. Alternatively, it can recognize and proactively address customer issues as they occur. For example, if a customer is having trouble loading a product into her virtual shopping cart, the system can offer immediate assistance that is more relevant to the issue than a general pop-up offering non-specific chat assistance.
Attorneys have to comb through large volumes of data to build their cases. The faster and more precisely an attorney can separate signal from noise, the more time she and her team can spend on litigation strategy. Historically, lawyers and their staffs have manually reviewed the documents, which can take weeks or months. Machine learning can speed the process and uncover important details humans may overlook.
"By combining machine learning and search analytics, you can find patterns in language that indicate peoples' behavior. Sometimes you're looking for the best example of a particular kind of behavior, or a particular incident," said Alexis Clark, solutions manager for data science at eDiscovery and information intelligence software provider Recommind, in an interview. "Sometimes you need to prove a pattern of behavior so you would want to find as many examples as possible."
Humans tend to look for familiar patterns, based on their experiences or other causes of bias. Machine learning can help yield accurate results, speed the process, and reduce related costs.
Cities around the world are using a crowd-sourced alternative to GoogleStreetView for city planning and to inventory roads and signage. Mapilliary uses machine learning to stitch together 3D visualizations of photos contributed by its more than 12,000 users. The images are available via an API.
In 2014, a city in Sweden reduced the speed limit in the center of town from 50km/h to 40km/h, which necessitated the updating of all relevant signs. It uploaded its old imagery to Mapillary, which was later used to verify the sign changes at each location.
"We're iterating based on human input. When users upload photos, it teaches us. We feed that into the next training of the system and it jumps up another level of performance," said Jan Erik, Solem, CEO and co-founder of Mapillary, in an interview.
Israeli communication services provider Orange (aka Partner) has been using machine learning for the past two years to help protect its business and customer data. The company was previously using an Information Detection System (IDS) and a Security and Event Management (SIEM) system, but the IDS system was producing approximately 800 alerts per day. With the LightCyber Magna active breach detection system, its security team can more effectively manage breaches and suspicious activity.
The system monitors all traffic coming from and being exchanged among PCs and servers, combinations of those things and more to identify anomalous behavior and to minimize its impact. Recently, the system detected malicious code in a video file that an employee had downloaded. The security team instantly notified the employee.
"We are in a very good position to know what users are doing that can affect the network and increase risk," said Arieh Shalam, CISO at Orange, in an interview. Wisely, the company supplements its technological improvements with education about such things as phishing and social engineering
Israeli communication services provider Orange (aka Partner) has been using machine learning for the past two years to help protect its business and customer data. The company was previously using an Information Detection System (IDS) and a Security and Event Management (SIEM) system, but the IDS system was producing approximately 800 alerts per day. With the LightCyber Magna active breach detection system, its security team can more effectively manage breaches and suspicious activity.
The system monitors all traffic coming from and being exchanged among PCs and servers, combinations of those things and more to identify anomalous behavior and to minimize its impact. Recently, the system detected malicious code in a video file that an employee had downloaded. The security team instantly notified the employee.
"We are in a very good position to know what users are doing that can affect the network and increase risk," said Arieh Shalam, CISO at Orange, in an interview. Wisely, the company supplements its technological improvements with education about such things as phishing and social engineering
-
About the Author(s)
You May Also Like