Qlik AutoML Models

Every Friday at Bitmetric we’re posting a new Qlik certification practice question to our LinkedIn company page. Last Friday we asked the following Qlik Data Architect certification practice question about Qlik AutoML models.

Judging by the amount of answers the AutoML function might not be so popular or well known yet, however:

The correct answer is D

Qlik AutoML

Qlik AutoML is a Qlik Sense SaaS specific capability which let you harness the power of machine learning on your datasets. It is a simple, no-code interface which you can use without (much) specific machine learning knowledge.

Currently Qlik offers three different models to use as basis for your predictions.

Regression

A regression model makes predictions on numbers in the future. Related questions which can be answered by this are:

• What is the expected revenue in Q2?

• How long will a patient be in the hospital?

• How many pieces of an article will be sold?

Binary Classification

The binary classification will give answers to yes or no questions. For example:

• Will my inventory run out?

• Will this customer churn?

• Does this car need preventative maintenance?

• Will this project be on budget?

Multi Classification

The multi classification answers questions where multiple outcomes are possible. Questions related to this model are:

• Which insurance plan is most fitting for this customer?

• Which inventory groups are most likely to be sold the most?

• Which product will a consumer buy?

How to use AutoML?

In order to properly use Qlik AutoML there are a few requisites you will need. First of, it is only available on Qlik Sense SaaS. If you have Qlik Sense SaaS available you will need to create two datasets. One training dataset and an apply dataset.

Training Data

The training dataset contains the old data in which your answer already has been answered. For example; if we would like to determine which insurance plan we need for a customer, we will need a dataset of customers and their chosen insurance plans. Next to this you will need to add relevant information to feed the AutoML algorithm with. So in this case: the insurance type, the cost of the insurance, the income of the customer, the value related to the insurance, etc, etc.

Apply Data

Next Qlik AutoML uses this training dataset to base it’s algorithm on. By determining the target it will know how to evaluate all the other fields and use these to base it’s predictions on. Now you will use this model on your current (production) dataset, which we call apply dataset. Based on the information in the model Qlik will now generate predictions based on the target column chosen. As we have learned this can be a fixed number (regression), a yes or no answer (binary) or multiple answers (multi class).

Things to keep in mind

• Limitations on the datasize based on your subscription model. As of writing the included-tier has a 100k cell limit.

• Data quality. Machine learning is a powerfull tool, but not infallible. A proper training dataset without clutter or NULL values is needed.

That’s it for this week. See you next Friday?