how to reopen results of an auto model after closing Rapidminer

1338773patti

as the title says.
我想开我从Aut的结果o model. I ran a Deep learning model. got some brilliant results, then saved the process, but how do I save the results to reopen them later??? Do i have to rerun the complete process again??

Sorry if this topic already excisted I can't find it anywhere.

rfuentealba

Hi@1338773patti,

I don't know a way to save results. There are some things you can do so you can reuse your training methods, though:

1. Split your process in two, maybe three.
   1. The first one, namely01 Trainingcan perform everything up to validation. Instead ofApply Model, though, you should use aStoreparameter to store the result of your algorithm, though.
   2. The second one, namely02 Runningneeds toRetrievethe result of your algorithm, and that should be connected to theApply Modeloperator.
2. Run your01 Trainingprocess everytime your training data changes.
3. Run your02 Runningprocess everytime you have new data (to process it against your latest trained model).

You can virtually store anything you want using theStoreoperator, as long as it can be converted internally into an IOObject, so you might want to place aStoreoperator just before any endpoint in your process, like here:

Now, don't be scared by IOObjects. These are the structures used internally by RapidMiner to store data, functions, models, etc. Just remember that these can beStore'd andRetrieve'd.

Here is a demo, usingRapidMiner Studioand the classic Titanic Dataset with a Deep Learning algorithm that saves results before displaying.

<描述一致=“中心”颜色= c“透明”olored="false" width="126">Load data.



<操作符= " true " class = " select_subproces激活s" compatibility="9.0.002" expanded="true" height="82" name="Define Target?" origin="EXPORTED_AUTOMODEL" width="90" x="45" y="34">













<描述一致=“中心”颜色= c“透明”olored="false" width="126">Define the target column for the predictive model.


<连接from_op = "定义目标”from_port = " example set output" to_port="output 1"/>





<描述一致=“中心”颜色= c“透明”olored="false" width="126">Should define a target column?

<操作符= " true " class = " select_subproces激活s" compatibility="9.0.002" expanded="true" height="82" name="Should Discretize?" origin="EXPORTED_AUTOMODEL" width="90" x="179" y="34">













<描述一致=“中心”颜色= c“透明”olored="false" width="126">Discretize by binning (same range per bin).














<描述一致=“中心”颜色= c“透明”olored="false" width="126">Discretize by frequency (same count per bin).








<描述一致=“中心”颜色= c“透明”olored="false" width="126">Should discretize numerical target column?

<操作符= " true " class = " select_subproces激活s" compatibility="9.0.002" expanded="true" height="82" name="Map Values?" origin="EXPORTED_AUTOMODEL" width="90" x="313" y="34">













<描述一致=“中心”颜色= c“透明”olored="false" width="126">Map some nominal target values to new values.








<描述一致=“中心”颜色= c“透明”olored="false" width="126">Should map nominal values?

<操作符= " true " class = " select_subproces激活s" compatibility="9.0.002" expanded="true" height="82" name="Positive Class?" origin="EXPORTED_AUTOMODEL" width="90" x="447" y="34">













<描述一致=“中心”颜色= c“透明”olored="false" width="126">Make sure that target is binary for positive class mapping.







<描述一致=“中心”颜色= c“透明”olored="false" width="126">Potentially define which one should be the positive class.









<描述一致=“中心”颜色= c“透明”olored="false" width="126">Should define positive class?

<操作符= " true " class = " select_subproces激活s" compatibility="9.0.002" expanded="true" height="82" name="Remove Columns?" origin="EXPORTED_AUTOMODEL" width="90" x="581" y="34">









<操作符= " true " class = " select_attribute激活s" compatibility="9.0.002" expanded="true" height="82" name="Remove Columns" origin="EXPORTED_AUTOMODEL" width="90" x="45" y="34">




<描述一致=“中心”颜色= c“透明”olored="false" width="126">Potentially remove columns.








<描述一致=“中心”颜色= c“透明”olored="false" width="126">Should remove columns?



<操作符= " true " class = " select_attribute激活s" compatibility="9.0.002" expanded="true" height="82" name="Remove Dates" origin="EXPORTED_AUTOMODEL" width="90" x="45" y="34">



<描述一致=“中心”颜色= c“透明”olored="false" width="126">Remove all date columns.




<描述一致=“中心”颜色= c“透明”olored="false" width="126">Transform all nominal columns to text so that we make sure that all will have polynominal type after the next transformation.




<描述一致=“中心”颜色= c“透明”olored="false" width="126">Transform all text columns into polynominal columns.






<描述一致=“中心”颜色= c“透明”olored="false" width="126">Turn all numerical columns (not integers though) into real columns.











<描述一致=“中心”颜色= c“透明”olored="false" width="126">Unify all value types













<描述一致=“中心”颜色= c“透明”olored="false" width="126">All general preprocessing steps happen inside this operator - double click on it to see the details.







<描述一致=“中心”颜色= c“透明”olored="false" width="126">Add a dummy nominal attribute to make sure that the loop will always deliver a result.













<描述一致=“中心”颜色= c“透明”olored="false" width="126">Calculate the number of missing values for this nominal attribute.




















<描述一致=“中心”颜色= c“透明”olored="false" width="126">Replace nominal missings with the word 'missing'.









<描述一致=“中心”颜色= c“透明”olored="false" width="126">Only replace missings if there are actually any missings.









<描述一致=“中心”颜色= c“透明”olored="false" width="126">Loop over all nominal attributes.

<操作符= " true " class = " select_attribute激活s" compatibility="9.0.002" expanded="true" height="82" name="Remove Dummy" origin="EXPORTED_AUTOMODEL" width="90" x="313" y="34">



<描述一致=“中心”颜色= c“透明”olored="false" width="126">Remove dummy attribute again.






<描述一致=“中心”颜色= c“透明”olored="false" width="126">Replace positive infinity values by missing.







<描述一致=“中心”颜色= c“透明”olored="false" width="126">Replace negative infinity values by missing.





<描述一致=“中心”颜色= c“透明”olored="false" width="126">Replace numerical missings with the average of the column.













<描述一致=“中心”颜色= c“透明”olored="false" width="126">Replace missing values.



<描述一致=“中心”颜色= c“透明”olored="false" width="126">Order columns alphabetically.




<描述一致=“中心”颜色= c“透明”olored="false" width="126">Model on cases with label value, apply the model on cases with a missing for the target column.



<描述一致=“中心”颜色= c“透明”olored="false" width="126">Sample down to 250,000 examples in case there are more.







<描述一致=“中心”颜色= c“透明”olored="false" width="126">Split of a validation set.











<描述一致=“中心”颜色= c“透明”olored="false" width="126">Train model.


<描述一致=“中心”颜色= c“透明”olored="false" width="126">Copy training data.


<描述一致=“中心”颜色= c“透明”olored="false" width="126">Copy validation data.


<描述一致=“中心”颜色= c“透明”olored="false" width="126">Create model simulator.





<描述一致=“中心”颜色= c“透明”olored="false" width="126">Copy model.



<描述一致=“中心”颜色= c“透明”olored="false" width="126">Apply model on validation set.













<描述一致=“中心”颜色= c“透明”olored="false" width="126">Performance on validation set.





<描述一致=“中心”颜色= c“透明”olored="false" width="126">Create predictions for cases without value and add explanations for predictions.



<描述一致=“中心”颜色= c“透明”olored="false" width="126">Create lift chart.











































Results:<br>1. Model simulator<br>2. Performance from validation set (split off before modeling)<br>3. Model<br>4. Predicted data with explanation viz (only if the data had missing labels)<br>5. Predicted data with explanation table (only if the data had missing labels)<br>6. Lift chart

Hope this helps,

Rodrigo.

1338773patti

首先,谢谢你的快速的交货tensive reply,

I do know you can rerun the process but that takes almost as long as first auto modeling.
so I was looking for a way to save the results so I can, when I close and reopen rapidminer, work on the simulation and the optimization again.
还您看如何将结果保存到使用in a presentation or a report.
I am a student and have to include every little step of the process in my reports.

rfuentealba

• 首先,谢谢你的快速的交货tensive reply

My pleasure. We are here to help!

• I do know you can rerun the process but that takes almost as long as first auto modeling.

Unfortunately yes, but that's the best use case for most of us.

• so I was looking for a way to save the results so I can, when I close and reopen rapidminer, work on the simulation and the optimization again.

You can save the simulation and the optimized work, that was what I suggested in the first place.

• Also I was looking how to save the results to use in a presentation or a report. I am a student and have to include every little step of the process in my reports.

I have three suggestions here, both of these require that you run yourAuto Model,Open the Processand save it at the end of your run. The idea behindAuto Modelis that you can accelerate the process of running your models, but you are then encouraged to modify it by yourself to suit your needs.

Take one of these:

1. Easy method: document each step by adding quick notes straight on the RapidMiner repository. You can then share a large screenshot.
2. Less easy method: record a video using Camtasia. No, just kidding. Add breakpoints before and after every important/interesting step you want to document, so you can build screenshots on intermediate results. To do so, you can just do a secondary click on each box.
3. Hard (using storage) method: instead of putting breakpoints, use theStoreoperator as I already suggested, your results will be readily available on a click. Just make sure you create aResultsfolder in your Local Repository and save your results with a different name on that folder. To work with the simulator, you can then go to the Local Repository, to the Results folder and double click the simulator.

Again, take a look at the XML process that I have shared with you. You can create a newProcess, open the XML panel inView > Show Panel > XML, paste the entire XML result and click on the green tick on the top, be back at theProcesspanel and your process will run seamlessly if you have RapidMiner 9. Look at the end of the process, where all thoseStoreoperators are. These will save data for you on the root of yourLocal Repository, so if you run it, you can see how did I do that.

All the best,

Rodrigo.

sgenzer

this sounds like an Auto Model feature request to me...copying@IngoRM

Scott

figueroajc

Yes, I ran into this same issue. Automodel took a couple of hours to run and generate a nice comparison between several different algorithms, then I found out I lost the comparison results since they are not saved anywhere and the export button only exports small bits and pieces of it at a time