# Statistical learning

## What is statistical learning and what can it be used for?

Statistical learning uses a variety of techniques to understand and model complex data sets. These data sets can consist of, e.g., sensor data, customer data or sales figures. The basic objective of statistical learning is to recognize the relationships between variables from observations. Relatively simple methods such as linear regression, nearest-neighbor analysis, but also more complex ones such as support vector machines or neural networks can be used. The fundamental challenge of statistical learning is the ability to generalize the knowledge gained, which makes it possible to obtain reliable information even from unknown data.

Basically, a distinction is made between so-called supervised and unsupervised methods. A supervised method is used to, e.g., predict the probability, e.g., whether an error in the production of a component occurs when measured sensor data are in a certain range. This means that a number of input variables, i.e., the sensor data, are used to predict a different parameter, in this case the occurrence of an error. In an unsupervised procedure, on the other hand, no quantity is predicted; only input quantities are examined for their mutual relationships and their structure. Both methods are often used to extract very useful information that cannot be found with the naked eye and superficial analyses.

Two main objectives are in the foreground: On the one hand, statistical learning can be used to make predictions and, on the other hand, to draw conclusions about the most important influencing factors of a process. Once you have found out which parameters influence the result, you can try to optimize these parameters, e.g., in the production process. The art of statistical learning consists of choosing the right method and the correct validation of the developed models. This ensures that both the complexity of the data is correctly captured and that the model delivers reliable results for unknown data.

Practically every company possesses data whose evaluation leads to optimized production processes, more efficient customer contacts, a stronger differentiation of its own products and thus ultimately to sound decisions and a stronger competitive position. Data is like natural resources; they have to be excavated and refined in order to realize their value. In the future, companies can be expected to differentiate themselves to a large extent by how much value they can derive from their data.

If you want to find out more about statistical learning or wonder how you can benefit from it in your company, we look forward to receiving your enquiries.