The concept of machine learning is that computer programs, when exposed to new data, can learn, grow, change and develop by themselves.

You can be forgiven for thinking the above statement belongs in the realms of science fiction – but if we look closely, smart machines and applications are steadily becoming a daily phenomenon. Here are a few machine learning examples we’re probably using (or have heard of) include:

  • Online recommendations: online data interactions to predict what users may want to buy, watch, or listen to.
  • Commercial flights: autopilot systems that take over from pilot shortly after taking off
  • Social media: algorithms for facial recognition
  • Mobile: speech recognition for voice to text or Smart Personal Assistants
  • Car navigation system: machine learning algorithms analysing speed of traffic/location data, and so on to suggest fastest routes and arrival times

As can be seen, machine learning has slowly evolved to be a part of our lives and represents a huge opportunity for organizations to become more imaginative and to look for further value from the vast data they have access to.

Traditional Data Analysis and Machine Learning

The key ingredient in the above examples is data —which is only going to grow and increase in size, volume, and complexity as the physical and cyber worlds interact. While traditional data analysis is great at explaining static data based on a model built on past data to establish a relationship between the variables, machine learning starts with the goal or outcome variables (like saving energy) and then looks for the variables and interactions.

Getting Started with Machine Learning

With all this in mind, the following check boxes will help guide your journey.

  1. Assess organizations data maturity: To begin an honest discussion about value creation, it’s imperative to do an audit on data governance, data warehousing, and overall data hygiene (look at sources, volume, third party data, and so on). It is vital for organizations to understand what data they have, what they don’t have, and what data they need to acquire.
  2. Define a clear problem that needs solving: Machine learning starts with the measurable results or outcomes we want to understand. As a side note, determine if the prediction you are trying to make (or decision you are trying to make) is complex enough to warrant machine learning. For example, are we dealing with a large volume of data and lots of variables?
  3. Assemble a diverse team: The best approach is to have a mix of technical experts who can deploy and maintain technical ecosystems, and data experts who can extract and translate data assets, along with functional or domain expertise to provide the business context.
  4. Embrace a ‘research’ type mindset: For machine learning to succeed, it’s beneficial for project teams to adopt a flexible and agile mindset to assess if an alternate approach is warranted. Be prepared to test, learn, and adapt (and repeat cycle). Is there a need to look at procure new data, adjust algorithms or go away from an existing model that may be showing bias?

In summary, it can be seen that machine learning is resource and time intensive. With increasing data available, machine learning gives businesses more options and opportunities. However, like any new journey, the above checkpoints can help to determine if you’re ready to take the next step.