From connected cows to everlasting elevators: How businesses are using machine learning
Connected cows taking part in the trial in Japan.
Image: Fujitsu/MTN/Microsoft
There's useful information lurking inside every business - whether it's a farmer knowing when a cow is ready to breed or a retailer knowing when you're likely to buy a soda.
But getting that information can require sifting vast quantities of data, all of which needs to be filtered and translated into helpful insights. That's where machine learning comes in.
Machine learning is the process of teaching a computer to impose structure and meaning on data.
When creating systems to carry out machine learning there are common methods for cleaning, linking and analysing data, and Microsoft recently launched a platform that offers services to help with these tasks.
Based on the Azure cloud, Microsoft's service enables firms to build machine-learning apps using tools that tidy up data and feed it into AI algorithms employed by products such as Bing search and the Xbox recommendation engine.
These applications can be written in the R and Python programming languages, or by using drag-and-drop tools and drop-down menus in a browser-based GUI.
Users can share their machine learning applications in a marketplace, with more than 30 currently available, ranging from customer churn prediction to text analytics.
"You can just go to that marketplace, subscribe to that API and integrate that into your website with just a small number of lines of code," said Joseph Sirosh, corporate VP for machine learning at Microsoft.
"What we've created is an app store-like experience on the cloud for machine learning APIs."
Applications can be deployed on the platform within minutes, according to Microsoft, as well as being updated and redeployed.
Sirosh hopes the service will lower the skills needed for companies to begin using machine learning.
"A small number of applications got created because of all that heavy lifting. Only a few companies were able to leverage most of the value. This democratises machine learning, makes it reachable."
The system has been available in preview since July last year and Microsoft has refined the service based on feedback from tens of thousands of customers.
Microsoft has been criticised for not integrating the service with competing cloud platforms so it can easily pull data from non-Microsoft sources, such as Google BigQuery and Amazon Relational Database Service. Sirosh said people can still create their own modules to pull data from sources of their choosing and that Microsoft is working on integrating new sources.
So just what are customers doing on the platform? Here are three examples of how businesses around the world are using machine learning.
The connected cow
Breeding cows can be tricky. The window for successful insemination is narrow - 12 - 18 hours every 21 days - and spotting it can require farmers to monitor tens or even hundreds of cows.
In Japan, dairy farmers employed a high-tech solution to noticing when a cow was getting frisky.
Eleven farmers fitted cows with internet-connected pedometers to report the number of steps they took each day to an Azure machine learning system. The system was trained to watch how the cows were moving and spot the spike in steps when the cow went into heat. Farmers would then be alerted by text, allowing them to artificially inseminate the animals at the optimal time.
The system proved 95 percent accurate in detecting the onset of ovulation and the number of calves born across the farms rose by an average of 12 percent. Farmers also reported having more time as they no longer had to watch for the signs themselves.
Researchers from Fujitsu, Microsoft's partner on the project, also discovered a link between when a cow's egg was fertilised and the sex of the calf. An animal conceived during the first half of the window for insemination would more likely be female and during the latter half would probably be male. This finding helped farmers control the number of cows and bulls in their herds.
The machine learning system was also used to spot health problems - with Fujitsu claiming it was able to detect about eight different cow diseases from the pattern of footsteps.
Keeping elevators running
German-based ThyssenKrupp looks after more than 1.1 million elevators worldwide and runs lifts in iconic buildings such as the One World Trade Center in New York and the 1263-foot CMA Tower in Riyadh, Saudi Arabia -- the country's tallest skyscraper.
Keeping those elevators running is a full-time business, and since last year the company has been working with Microsoft to build a monitoring system that feeds data from its elevators and escalators to the Azure cloud platform, using Microsoft's Intelligent Systems Service to help capture that information and its Machine Learning service to make sense of it.
The aim is to develop a system that knows what repairs need to be carried out before anything breaks and which can advise engineers on what work needs doing during call-outs.
By monitoring information from its lifts ThyssenKrupp plans to target when and where it carries out maintenance. Rather than scheduling a routine service every few months, the frequency and nature of these services would instead be based on how each elevator is functioning. Keeping tabs on their workings will be the Azure machine learning service, which will monitor details such as how often a lift door opens or the energy expended to drive the elevator.
ThyssenKrupp and Microsoft have built an expert system based on Azure Machine Learning that can tell an engineer what's likely to be wrong with an elevator during a service visit -- listing the four most probable causes of an error code and ways to test for each problem.
Feedback from engineers on what the actual problem was and how they fixed it will help the system learn and refine how it interprets such error codes and advises other engineers in future.
ThyssenKrupp has trialled a proof-of-concept version of the system on less than 50 elevators and plans to begin rolling it out in earnest this year, expecting to extend it to about 600,000 elevators and escalators in 12 months' time. Eventually the company intends to use the system to monitor about 60 percent of its elevators worldwide.
Predicting purchases
JJ Foods is a major wholesale retailer in the UK, delivering thousands of different foods and household goods to more than 60,000 customers across the country.
The firm wanted to anticipate customer needs, and worked with Microsoft to develop a system that could predict what returning consumers would buy, in order to fill their shopping cart in advance.
JJ Foods' application captures customer behaviour online, sends it to the Azure ML service and generates a list of what customers would likely purchase in future.
The system was trained using three years of transactional data and integrated with the firm's Microsoft Dynamics AX system, an ERP system that handles information from across the business.
The system, which took about three months to implement, predicts purchases for customers buying online or over the phone. These forecasts have proven to be pretty accurate, in general forecasting about 80 percent of the goods a customer will buy.
The software also analyses purchases and tries to predict other products that might be needed, for example cooking oil for a restaurant ordering meat and vegetables.
The company is hoping the system will both drive sales, improve customer satisfaction and help it stock its warehouses more effectively.
