- Regression


Regression as parameter prediction is a broad concept that covers everything from predicting house prices, to predicting resource consumption or predicting how much weight concrete can carry. In short, everything where you can put something in a box, or mathematically speaking, all that is continuous.

A common factor for all these cases, however, is that the underlying mathematics that an AI / ML regression model will use to make the prediction is in the majority of cases the same. Thus, the same basic architecture can be used to make many different regression models.

We are experts in the whole process regarding regression models, everything from development to implementation and maintenance throughout the model’s life cycle and can, in addition to setting up the model, also advise you how to further improve the model’s result through further data collection.

Requirements for regression models

When creating a regression model, it is important that you already have data to base the model on, but the amount required is less than what you probably think. If what you have now shows potential in creating a model, it is easier to make a POC/proof of concept with what you already have and then improve it later, because we can always collect further data to improve the model’s predictions if the foundation appears promising.

It is important that the data is in a structure that the model can read, because ML models are very sensitive to how the data is provided. We can help you solve this problem by using data augmentation to ensure that the data provided to your model has the correct structure.

In addition, it is important that you know in advance what it is you would like to have classified, so that we can point the model in the right direction. If there is any doubt about this, we can also help you determine the scope for the model and prepare the business case through an AI / ML clarification.

A parameter regression model's technical possibilities

With a regression model, you can predict certain values that are critical to your business. In addition, you will also receive regular feedback on the model's performance, so you know how much you can trust it.

It is necessary to consider the entire infrastructure regarding the model when generating a solution. This is because the model alone is not that useful if you do not know how to give it the correct type and structure of data, or how to return a result from the model. We can help set up the entire infrastructure so that the model can be integrated into your system without you having to think about the maintenance of the model.

The process regarding implementing regression models

Q-analytic’s process when making regression models is:

  • Initial meeting
  • Workshop
  • Data is accessed
  • Model generation
  • Result / Delivery
  • Maintenance

Initial meeting

At the introductory meeting we will talk about the considerations you have and what it is you would like to have regressed. We will provide input and advice along the way so we can understand your needs and narrow down to the solution that will be optimal for you. We will also talk about how the model should be integrated into your systems, as this should be taken into account from the start.


For the workshop, we focus on the technical aspects of model generation, as well as the integration of this in your systems, here we have already determined the scope and are now concentrating on how we make it possible. After the initial preparations are we to be presented with what your data includes so we can see what it is we need to work on. And get a review of it so that any questions we may have about this can be clarified.

Data is accessed

Here we get access to the data so that we can start generating the model per your needs. Normally we will do the data processing in our own office, but if your data is extra sensitive so that it can not leave your premises, we can also easily agree that we perform the model generation with you observing our work.

Model generation

We generate the model based on the instructions we agreed on for the workshop. We will also produce graphs and reports that describe how the model performs so you have full insight into what you get delivered.

Result / Delivery

We present the result to you and hand over the regression model, and we will help you integrate it into your systems if need be.