Automodel meanAbsolute rootMeanSquared error and other missing algorithms

ozcan

Hi,
1) Im trying to find meanAbsolute rootMeanSquared error in Automodel results. But these values are not available in auto model results ?
2) Part, Bayesian Network Algorithm are available in rapidminer ? like weka.
3) Boosting, bagging, KNN Algorithm are available in auto model ?
4) What is deep learning below neural nets folder ? Is it multilayer perceptron ? If it is not, Is there a multilayer perceptron in Rapidminer? 

varunm1

Hello @ozcan

1) Im trying to find meanAbsolute rootMeanSquared error in Automodel results. But these values are not available in auto model results ?

Are you asking this while trying to solve the regression problem in auto model? If so, then there is an absolute error and RMSE displayed in the results of auto model. 

Auto model displays limited performance metrics. If you want to look at different performances, you can always open automodel process by clicking "Open process" and go to "Validate Model" and inside "Performance hold out set" you can see the performance operator and select other performance metrics.

3. Boosting, bagging, KNN Algorithm are available in auto model ?

Algorithms related to boosting (GBT) and bagging (Random forest) are available. One thing to be noted here is auto model selects algorithms automatically based on problem types and data properties (maybe some secret algorithm which @IngoRM knows  ). If we want to apply more models, we need to open any auto model process and customized them for whatever algorithm we want to use. 

4. What is deep learning below neural nets folder ? Is it multilayer perceptron ? If it is not, Is there a multilayer perceptron in Rapidminer?

Yep, it's a feed-forward neural network that can have multiple layers. You also have AMLP which is an auto multi-layer perceptron algorithm and "perceptron" that is a single perceptron. The general neural network operator is also of the same category. The major difference between the deep learning and neural networks operator in rapidminer is the advanced options present in deep learning operators. You can go through the help text in the rapidminer studio to get more information.

2) Part, Bayesian Network Algorithm are available in rapidminer ? like weka.

Rapidminer has Weka extension that can be installed from the market place and it has bayesian based algorithms that you might be referring to. I am not sure if there are any native operators in RM other than naive bayes and bayesian boosting technique.

Hope this helps. Please let us know if you need more information.

[Deleted User]

Hello
@ozcan

This link is for Auto model

https://docs.rapidminer.com/latest/studio/guided/auto-model/

All the best
mbs

varunm1

Hello @ozcan

Yes, please see the definition in RM below.

The absolute error is calculated by adding the difference of all the predicted values from actual values of the label attribute, and dividing this sum by the total number of predictions.

MartinLiebig

Hi @ozcan,

yes, absolute error is the mean of all absolute errors in your predictions.

Best,
Martin

varunm1

Hello @ozcan

1) Im trying to find meanAbsolute rootMeanSquared error in Automodel results. But these values are not available in auto model results ?

Are you asking this while trying to solve the regression problem in auto model? If so, then there is an absolute error and RMSE displayed in the results of auto model. 

Auto model displays limited performance metrics. If you want to look at different performances, you can always open automodel process by clicking "Open process" and go to "Validate Model" and inside "Performance hold out set" you can see the performance operator and select other performance metrics.

3. Boosting, bagging, KNN Algorithm are available in auto model ?

Algorithms related to boosting (GBT) and bagging (Random forest) are available. One thing to be noted here is auto model selects algorithms automatically based on problem types and data properties (maybe some secret algorithm which @IngoRM knows  ). If we want to apply more models, we need to open any auto model process and customized them for whatever algorithm we want to use. 

4. What is deep learning below neural nets folder ? Is it multilayer perceptron ? If it is not, Is there a multilayer perceptron in Rapidminer?

Yep, it's a feed-forward neural network that can have multiple layers. You also have AMLP which is an auto multi-layer perceptron algorithm and "perceptron" that is a single perceptron. The general neural network operator is also of the same category. The major difference between the deep learning and neural networks operator in rapidminer is the advanced options present in deep learning operators. You can go through the help text in the rapidminer studio to get more information.

2) Part, Bayesian Network Algorithm are available in rapidminer ? like weka.

Rapidminer has Weka extension that can be installed from the market place and it has bayesian based algorithms that you might be referring to. I am not sure if there are any native operators in RM other than naive bayes and bayesian boosting technique.

Hope this helps. Please let us know if you need more information.

[Deleted User]

Hello
@ozcan

This link is for Auto model

https://docs.rapidminer.com/latest/studio/guided/auto-model/

All the best
mbs

ozcan

Thanks for reply, really helpful . One more question , Im trying to find mean absolute error. There is a absolute error in performance results .I didnt see mean absolute error in performance metric. Can I evaluate absolute error as mean absolute error?

varunm1

Hello @ozcan

Yes, please see the definition in RM below.

The absolute error is calculated by adding the difference of all the predicted values from actual values of the label attribute, and dividing this sum by the total number of predictions.

MartinLiebig

Hi @ozcan,

yes, absolute error is the mean of all absolute errors in your predictions.

Best,
Martin