Machine Learning

A frequently used use case for ML and AI is recommendation systems, where algorithms can come up with different suggestions based on previous data. This is most often used in films, books, experiences and e-commerce products with a view to generating additional sales. However, it can also be used more internally, for example when proposing treatments for patients based on their symptoms.

It is often useful to be able to predict the future, e.g. in connection with resource planning. Unfortunately, predicting the future is no mean feat, especially if there are several parameters that affect one another. In such situations, ML can be a useful tool for uncovering complex relationships and subsequently using them in connection with planning and resource optimisation.

Forecasting can, for example, be used as a tool in Facility Management, where, among other things, interaction is created between route planning of service tasks, energy consumption, personnel and other resources.

Many e-commerce websites store large amounts of data about customers, their behaviour and their purchases. This data can be used to learn more about customer segments. By learning more about the customer segment, marketing can, among other things, be targeted at this particular segment.

Identifying deviations in real time can be particularly relevant in monitoring heavy processes, where it is not possible to spot deviations through standard monitoring. With a sufficiently large amount of data, it is possible to identify what is “normal” and thus also what is not. The method is called Abnormality Detection. This takes place without any previous definition of what is normal.

In most applications, dependent parameters/variables emerge that are desirable to optimise on. These might include power consumption, time, performance or quality. In such applications, agents can be created with the help of ML to control, optimise and keep track of the desired parameters. The application can then be simulated to identify the optimal setting for optimisation of one or more parameters.

With predictive maintenance, data-driven predictions can be used to optimise maintenance and thus extend the life of the product while minimising the risk of breakdowns. Diverse data sources are often used to make the desired prediction. This might be changes in energy consumption, the number of revolutions, sound, rate of utilisation as well as input from various sensors. Ultimately, the goal is to provide maintenance or service in a timely manner.

ML can be used to automate simple processes, such as filtering and sorting data. This can be anything from documents, emails, reviews and results from questionnaires.

It can be difficult to obtain the initial overview and identify relationships in large amounts of data. ML algorithms can be used here to cluster data in order to identify patterns and relationships that can help to provide greater insight and overview.

The technology can, for example, be used to automatically cluster large amounts of text to identify patterns, providing a better point of departure from which to understand the problem.

Clustering is just the first step in the task of understanding a problem and coming up with proposed solutions.