Using IBM Machine Learning to Help Solve Real World Business Problems

Billions of connected devices, zetabytes of data, power and brand loyalty now in the hands of the consumer, businesses having to market and sell to each and every one of us.  What just happened?  Three driving forces – mobile, cloud and a continuing explosion of data. Everything just got personal.

How can any business make sense of it all? How can they learn and avoid making the same mistakes – and become smarter. Oh – and did I mention much of this needs to happen in real time?

That’s where Machine Leaning as part of a cognitive strategy comes in to its own. Read my earlier blog on Enterprise-ready Machine Learning.

 

Machine Learning  – Get Smart.

It’s all about doing things smarter.  Regardless of industry, machine learning can greatly assist organizations in making progressively better decisions by constantly adapting and learning.

Below are a just a few examples of how organizations have leveraged machine learning capabilities to achieve significant business benefits or provide better service to their consumers and constituents.

 

Healthcare and the Zika Virus – a non government organization (NGO) healthcare based company needed to understand which primates are most likely reservoirs of the Zika virus and which mosquitoes are most likely to be the carriers.  IBM developed a new algorithm to help organizations identify the most likely animals and to help them with early detection, prevention and managing virus spread.   The benefits of course are very clear – by better understanding the disease the organization was able to proactively fight it before the epidemic occurs.

 

Government organization addresses long term unemployment – The problem of youth unemployment can be expressed as a statistic but can’t be solved as one. It needs to be solved one person at a time. A European public job agency believed that within its sprawling and complex base of employment data lay the indicators of which young job seekers are most at risk of long-term unemployment and, most importantly, why. Cracking these patterns would help lead to breakthroughs. The organization ran a decade’s worth of deep historical data through machine learning algorithms. The result was a predictive model that not only quantifies long-term unemployment risk but also breaks down the impact that each controllable variable – such as having or not having a driver’s license – has on the big picture. This enabled job counselors to offer job seekers guidance with an unprecedented level of detail, personalization and effectiveness. Traditionally they relied on their own experience and intuition of “what works”. The new solution applies machine learning analysis to the agency’s full record – 10 years’ worth of historical data – to show what really works. Counselors were then able to provide recommendations to job seekers with a high level of confidence. It set the stage for a transformation, both in the depth of its understanding of the youth unemployment issue and in its ability to bring personalized options to young job seekers. Because job counselors have access to granular, personalized insights into the “whys” of youth unemployment – insights surfaced by machine learning and NLP – they can make practical recommendations on closing skill-set and qualification gaps with confidence that it will help make a difference.

 

Automotive – Improving Vehicle uptime. A multi-national car manufacturing company based in Europe wanted to improve transport vehicle uptime and avoid unplanned stops so that it would achieve its vision of becoming the world leader in sustainable transport solutions. By building a predictive analytics platform the company gained the ability to identify the necessary parts and provide repair instructions, even before a truck arrives for service. This capability helped reduce diagnostic time by up to 70 percent and cut repair time by more than 20 percent .The client also gained the ability to plan maintenance better by performing preventive maintenance.

Further, the solution applied machine learning techniques to automatically discover patterns and learn from the vast amount of data it collected. Additionally, the company consolidated under one roof the people and systems needed to monitor and respond to vehicle issues in near real-time, including around-the-clock support. The process changes also helped the company maintain its commitment to core values of quality, safety and environmental care.

 

Oil and GasIncreases outage event detection accuracy by more than 95 percent  – For oil and gas companies, powerful and complex equipment such as compressors comprise the heartbeat of production. Keeping them running is a top priority. Companies use sophisticated asset surveillance systems to look for the signs of impending outages. Still, those systems are missing a significant share of outage events because the mix of warning indicators can be simply too complex to discern from the flood of noisy data generated by sensors. One company embraced a new way of measuring the health of its oil production assets. It’s using machine learning algorithms to automatically build complex, multivariate and far more flexible rules that define which changes in vibration patterns, pressure and the like are true anomalies. And because these machine learning algorithms are self-correcting, they get more accurate over time. The result: a quantum leap in outage detection accuracy.

The solution is game-changing for the company because the use of machine learning technology provides a far higher level of predictive accuracy than would be possible with the traditional statistical anomaly detection approaches used by oil companies. Because machine learning discovers complex pre-failure patterns within sensor readings, rather than just seeking out the simplified patterns that were predefined by engineering staff, the solution is far less likely to miss signs of impending equipment failure

 

Computer Services – digital image processing. A North American company collects more than two million square kilometers of earth imagery every day from high-resolution satellites and is used by urban planners to US and foreign defense and intelligence agencies. Its imagery is also used commercially for navigation technology and web mapping applications. The company developed a new machine learning system designed to help minimize the amount of manual review that is required to detect cloud cover in satellite images of Earth. The system consists of a set of sophisticated image-processing functions using an algorithm that effectively classifies each pixel within an image as either cloudy or non-cloudy by comparing it with the millions of pre-classified examples in the training data set. The company designed the models using IBM machine learning technology and then exported them to their custom-built system where they are compiled and deployed.

Over a test data set consisting of 600 images, comprising a mix of cloudy and non-cloudy images, the new system reported a true positive rate of 90 percent and an associated false positive rate of just six percent. For the non-cloudy scenes in the data set, which are the most valuable images from the client’s perspective, the system reported a false positive rate of just 0.1 percent, compared to 22.7 percent for its previous system. The company saw the false positive rate for its black and white satellite fall to 4.9 percent, from the previous rate of 71.1 percent.    They expect the system will grow ‘smarter’ over time and will hope to reduce the amount of images that require manual review by 90 percent.

 

UK based Retail Company focus on loyalty and customer satisfaction – When customers order online, they do so primarily for convenience and cost savings. Most are willing to forego the immediate satisfaction of trying out the item in person, paying for it immediately and walking out the door with the new product in hand. When online shoppers receive merchandise that fails to meet their needs or expectations, their disappointment and the effort required for resolution may discolor their impression of the transaction, eliminating both the convenience and the cost savings.  Studies demonstrate that 80 percent of first-time customers who return an order will never purchase from a company again. This retailer uses advanced analytics to identify both the products that are returned most frequently and the ineffective marketing tactics that elevate the number of returns. Using statistical and machine learning models, the company takes advantage of rapid alerts and automated information feeds on products that are returned so that it can quickly identify problems with sales and anticipate and help resolve evolving issues.

The solution allowed the retailer to develop a customized response system to automatically contact new customers who returned items, enabling customer service specialists to offer incentives to encourage disaffected buyers to try again. This personalized, near real-time response makes a tremendous difference to customer perceptions of retailers, enabling the company to help improve customer loyalty.

 

Energy and Utilities integrating renewable energy into the grid. The Vermont Electrical Power Company  (VELCO) worked with IBM Research to develop an integrated weather forecasting system to help deliver reliable, clean, affordable power to their consumers while integrating renewable energy into the grid.  The solution combines high resolution weather with multiple forecasting tools based on machine learning.  The machine learning models are trained on hindcasts of weather correlated to historical energy production and historical net demand.

The results are some of the most precise and accurate wind and solar generation forecasts in the world. This powerful tool turns multiple streams of data—transmission telemetry, distribution meter data, generation production, highly precise forecast models—into actionable information using leading edge analytics. A collaborative achievement involving dozens of in-state and regional partners and the formidable intellectual resources of IBM Research, VWAC’s results are significant and its value already demonstrated, even as further benefits continue to emerge. To find out more and the actual business results  watch the video on the VELCO website Courtesy Vermont Electrical Power Company web site and video.

 

Machine Learning – It’s all part of the Data Science Experience.

You don’t have to be a data scientist to use IBM machine learning. But even if you are we help automate the experience as much as possible with integrated tools that guide you through a step by step process.  Or simply use existing / prebuilt machine learning capabilities as a service (MLaaS).  There’s a lot happening. If you haven’t already done so I invite you to sign up for the data science experience at datascience.ibm.com.

 

For more information on IBM’s cognitive strategy and machine learning capabilities click this link : ibm.com/outthink

 

 

Dinesh Nirmal, 

Vice President, Development, Next Generation Platforms, Machine Learning, Big Data & Analytics 

Follow me on Twitter @DineshNirmalIBM

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s