Classification as parameter prediction is a broad concept that covers everything from suggestion algorithms which are seen at online shops, to diagnosing patients for the healthcare sector. In short, everything where you can put something in a box, or mathematically speaking, everything is discreet.

Common to all these cases is that the underlying mathematics that an AI / ML classification model will utilize to make the prediction is for the majority cases the same. Thus, the same basic architecture can be used to make many different classification models.

We are experts in the whole process regarding classification 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.

When creating a classification 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, as ML models are very sensitive to how the data is given to them. We can help you with this by utilizing data augmentation to ensure that the data given to your model has the right 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.

Technical possibilities with a parameter classification model

A classification model’s technical possibilities include the ability to predict discrete values or sort elements into the correct boxes. This will also provide you with ongoing feedback including key figures about how the model performs.

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.

Q-analytic’s process when making parameter classification 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 classified. 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.

Workshop

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 classification model, and we will help you integrate it into your systems if need be.

- Classification

Description

Requirements for parameter classification