machine learning, manufacturing, predictions

Yes, machines can learn too! Put simply, machine learning is the ability for computerized machines to learn how to make predictions and/or perform classifications without explicitly being taught or programmed. Just like humans have different learning styles and problem-solving techniques, so do machines, and there are numerous approaches to machine learning depending on the task at hand.

At a Glance

By creating smart solutions using machine learning algorithms it is possible to increase productivity and improve efficiency and accuracy in manufacturing environments. Imagine any type of prediction scenario. You might be making predictions on some of the following aspects of manufacturing:

  • when to schedule downtime for your factory machines and gadgets
  • how much scrap the production lines will produce at a given time
  • what your factory schedule will look like in order to meet seasonal demand during the holidays
  • which sales strategies to apply to your customers
  • whether or not your next promotional campaign will reach the right market and audience

The answers you seek almost always come from past experience and insights into current and future conditions. This past experience may have been gathered over a span of days, weeks or even years, and the current and future conditions may come from various market trends, your sensor data, your customer characteristics, etc. A variety of past, present and potential future data goes into making predictions. When humans make these predictions, there can be a level of uncertainty and bias to it. This same data can be fed into machine learning algorithms to create prediction models that not only predict outcomes with greater accuracy, but evolve and get more accurate over time as more factors and attributes are added to the knowledgebase for predictive outcomes.

How does Machine Learning work?

When I think of machine learning, I think about my experiences with Algebra in high school. Algebra math was tough to understand at first glance so I started by reading the chapter and going through many examples. I used a lot of trial and error, always checked my answer and practiced a lot of similar problems until I mastered them.

With machine learning you need a lot of good training data or examples to work with. Through trial and error, you can try various algorithms, verify the results with test data and try it on various data sets as practice. Once you have the right model, it can then be used for a variety of similar data sets. One distinctive advantage that machine learning has over a human approach is an unlimited appetite for data and virtually infinite capacity to churn on that data. This allows machine learning to be applied at a scope and scale of problems that would simply be impossible for a person to solve alone.

Just like there are various techniques for humans to learn, there are numerous algorithms and methodologies available for machine learning. Classification and regression are the most common categories, and each provides a different outcome. Fundamentally, classification is about predicting a label while regression is about predicting a quantity. Various algorithms have their pros and cons and depending on the type of data and expected outcome, there are one or more to pick from. Most require feeding a lot of historical data as examples to the algorithms, which requires managing and cleansing countless data often found in a data lake.

Looking Ahead

Machine learning is allowing businesses to optimize and accelerate the repeatable processes that are often time-consuming and inaccurate by nature. It is also enabling computers to assimilate vast amounts of data accumulated from various IIoT enabled equipment and make smart decisions and predictions, which would otherwise be too difficult or time consuming for humans.

At QAD Labs, we are exploring and experimenting with advanced technologies, like machine learning, to solve the problems of today and tomorrow. We are poised and ready for the industry disruptions that will create game-changing challenges for manufacturers and stand ready to assist in solving those that lie ahead with digitization, Industry 4.0 and beyond.

We are always looking for additional lab partners! Whether the activity is centered on machine learning, IoT, Robotic Process Automation or other emerging technology, QAD is providing a platform and resources to jointly move our customers’ capabilities along. If you have a need or an idea for experimentation around advancing technology, please let us know at