The Case for Machine Learning in Business (original) (raw)

Originally published in the ITS Ghaziabad 2nd CXO Meet Souvenir, "Digital India Mission: Transforming India for Tomorrow."

Achievements in machine learning are coming at ever-increasing rapidity over the past several months. You are likely familiar with the recent accomplishments associated with machine learning, especially those of so-called deep learning, or the use of multi-layered artificial neural networks. These specific achievements include the high profile AlphaGo and Deep Dream, along with numerous others in the realms of computer vision and natural language processing. Interestingly, a number of these recent mainstream successes are primarily attributable to Google in one form or another.

Machine learning is closely related to, and often confused with, artificial intelligence (AI). When we think of AI, we generally think of robotics, intelligent agents, and perhaps even absurd doomsday scenarios courtesy of Hollywood. We don't, generally, think of business cases involving general AI, or even leveraging more narrow automated learning to solve specific business problems. But that's exactly what machine learning in the enterprise can accomplish, and exactly why you should be considering it today.

Source: Gartner (August 2016)

Gartner’s 2016 Hype Cycle shows that machine learning is now in the “peak of inflated expectations,” having made its way out of the “trough of disillusionment,” which can be interpreted as meaning that its usefulness is now understood, even if harnessing this is not (yet).

The following explores 3 distinct machine learning avenues you can pursue right now to help develop and achieve your business goals.

Deep Learning for Business

Convolutional and recurrent neural networks have played important roles in recent state-of-the-art computer vision and Natural Language Processing (NLP) advances. These often sound like esoteric research goals, but these technologies have legitimate business applications, and can have profound affects on bottom lines, and can spawn entire industries.

Computer vision is making automated driving cars a reality. This is a huge industry set to explode in the next few years, and one which will generate billions of dollars in revenue. Computer vision is also responsible for the development of specialized applications such as eye, head, and gait tracking. These have practical applications in health and safety fields, and, aside from increasing the quality of life for individuals, their development and implementation will also drive an entirely new industry. Open source deep learning frameworks such as Google’s TensorFlow, when matched with the required hardware, can be used to build and train models which are able to perform these types of tasks.

NLP is even more universally-applicable to a variety of business types. Natural Language Understanding (NLU) involves computers actually understanding natural language and free form conversation and text. Such systems can help to make sense of vast amounts of text that an organization may have acquired, be it from social media interactions, customer emails, or any other body of text you may have available. TensorFlow, when paired with currently available pre-trained models such as Google’s SyntaxNet, provide powerful out-of-the-box implementations for organizations to easily begin taking advantage of these previously out of reach technologies.

From here, all sorts of additional processing on text can be performed. Sentiment analysis is a related technology which automatically gauges the attitude or emotion associated with written text, the results of which can be actionable. Think of identifying products which garner overwhelmingly negative reviews on social media, and being able to make improvements based on this data.

Social media analytics is also closely-related, and indeed overlapping, and can extend additional technologies into the realm of analyzing social data. In today’s technological economy, what business would not be able to make use of techniques such as these?

Automated Machine Learning

If, as Sebastian Raschka has described it, computer programming is about automation, and machine learning is "all about automating automation," then automated machine learning is "the automation of automating automation." Follow me, here: programming relieves us by managing rote tasks; machine learning allows computers to learn how to best perform these rote tasks; automated machine learning allows for computers to learn how to optimize the outcome of learning how to perform these rote actions.

This is a very powerful idea; while we previously have had to worry about tuning parameters and hyperparameters, automated machine learning systems can learn the best way to tune these for optimal outcomes by a number of different possible methods.

This sounds very theoretical, but it is most assuredly not. While previously organizations had to worry about having teams of highly knowledgeable data scientists and similar analytics specialists at their disposal to be able to even tinker with machine learning or data mining, such automated systems can alleviate much of this required know-how by sheer automation. Of course, interpretations of results still requires statistical savvy, and more complex problems are not for the unskilled, regardless of current technological assistance.

However, the important takeaway is this: with some basic technical understanding and some time investment, organizations can employ one of several open source systems like TPOT or Auto-sklearn to dabble in automated machine learning for specific problems and see what insights, if any, can be gleaned. If deeper analytics and statistical modeling are desired, especially anything that would be used for steering business practices, certainly employing qualified individuals to helm such automated systems is necessary; however, automated machine learning is able to lower the barrier of entry, and remove the potential uneasiness surrounding the prospect of, machine learning in the enterprise.

There has not been any time in the past where machine learning has been as accessible to business.

Application Programming Interfaces (APIs)

Like so much of contemporary technology, machine learning is no longer solely an in-house operation. Pairing locally- or cloud-hosted data with available machine learning Application Programming Interfaces (APIs) is becoming an attractive manner for which to pursue routine analytics within the organization. A wide variety of machine learning, data science, and cognitive computing tasks can be accomplished using such APIs.

Whether looking to make predictions based on “conventional” data, or to extract text from audio or video and perform subsequent analysis, there are numerous options now for such an approach.

Hewlett-Packard Enterprise (HPE) offer its Haven OnDemand machine learning APIs, which are made up of more than 60 such APIs for processing, connecting, and analyzing data in a variety of ways, including text extraction and a variety of other data science tasks. Microsoft Azure Cognitive Services is another option, offering a number of APIs for accomplishing similar tasks as well.

For more on the open API economy and how it has boosted analytics, see this article by Kaushik Pal of TechAlpine.

Conclusion

With everything taking place in machine learning these days, and given its ease of access and low barrier of entry, there is little reason not to be employing these technologies in your business right now. I encourage everyone to search out the best way in which to take advantage of machine learning, and wish you the best in your pursuits to do so.