Skip to main content
Announcements
Have questions about Qlik Connect? Join us live on April 10th, at 11 AM ET: SIGN UP NOW
cancel
Showing results for 
Search instead for 
Did you mean: 
AayushShah
Contributor
Contributor

How to perform Multi Target Regression (MTR) on Qlik?

Hello, I am new at this platform, In my use case I want to train a model which can predict multiple targets from the given features. Suppose my data has some features (X1...Xn) and 4 targets. (Y1,...Y4). 

I want to run a regression model that should predict all 4 targets by simply providing the features. So, I am looking for a way to do that on Qlik. Will anyone please help?

Now, there can be some relation between targets as well, thus creating separate models for every 4 targets might not result in the best model. Your guidance will help the most.

Seemingly, it is a Multi Target Regression problem. Will you please help me how to get this done?

Thank you,

Aayush Shah

1 Solution

Accepted Solutions
Kyle_Jourdan
Employee
Employee

This does help clarify, but you should still be fine creating individual models for each target. The other "targets" in the dataset can just become features in the respective models. Anytime you select your target, all the other columns remaining in the dataset become features (with some exclusions based on data quality), so if "Height" is your target, "Weight" becomes a feature, and vice-versa.

View solution in original post

4 Replies
Kyle_Jourdan
Employee
Employee

I assume you're referring to a single target with multiple labels (a single column with labels Y1,...,Y4)? If so, Qlik AutoML will solve this as a "multiclass" experiment.

If, however, you're referring to multiple target columns (Y1, Y2, Y3, etc. that contain "true/false"), you would need to build multiple binary classification experiments using the same dataset (and therefore feature set).

AayushShah
Contributor
Contributor
Author

Thanks, @Kyle_Jourdan for the reply. I think you understand my problem as a classification problem. Actually, it is a regression problem.

Suppose:

Age, BMI, GenderHeight, and Weight are in my dataset. From which I want to predict Height and Weight (as targets) from the features Age, BMI and Gender.

So, when I train the model it should learn relationships for my targets from the features and also the relationships between the targets.

As of your reply, I think we can create different models for each target but doing so, the models will not be able to learn the relationship between targets because when training the model1 for Height, the second target Weight will not be considered, and vice-versa.

So, is there any way to approach this task on Qilk? Or do we have to manually create separate models for each?

___

Additional context:

We can't include the rest of the targets in training like:

Suppose-Data: X1, X2, X3, Y1, Y2, Y3

Model1: Features: {X1, X2, X3, Y2, Y3}  | Target: {Y1}

Model2: Features: {X1, X2, X3, Y1, Y3}  | Target: {Y2}

Model3: Features: {X1, X2, X3, Y1, Y2}  | Target: {Y3}

The procedure given above will not work. Because while training, it requires the other targets to be in the feature-space. But while testing (real world prediction) we will not have them. Thus, we will not be able to predict in the real world.

 

I hope I made it clearer. I hope I could explain my use case. If there is anything that can be done, please let me know. 

Best,

Thank you.

Kyle_Jourdan
Employee
Employee

This does help clarify, but you should still be fine creating individual models for each target. The other "targets" in the dataset can just become features in the respective models. Anytime you select your target, all the other columns remaining in the dataset become features (with some exclusions based on data quality), so if "Height" is your target, "Weight" becomes a feature, and vice-versa.

Antonio365
Contributor
Contributor

MSE is calculated by: measuring the distance of the observed y-values from the predicted y-values at each value of x; squaring each of these distances; calculating the mean of each of the squared distances.


@AayushShah wrote:

Hello, I am new at this platform, In my use case I want to train a model which can predict multiple targets from the given features. Suppose my data has some features (X1...Xn) and 4 targets. (Y1,...Y4). 

I want to run a regression model that should predict all 4 targets by simply providing the features.Panorama Login Charter So, I am looking for a way to do that on Qlik. Will anyone please help?

Now, there can be some relation between targets as well, thus creating separate models for every 4 targets might not result in the best model. Your guidance will help the most.

Seemingly, it is a Multi Target Regression problem. Will you please help me how to get this done?

Thank you,

Aayush Shah