Am I building my Deep Learning model right?

jakob_roetner

I'm building a Deep Learning model in Rapidminer right now and I have some basic knowledge in Machine Learning, but I'm sometimes a bit confused how to implement my ideas in Rapidminer.

The basic idea is this:

• I have a dataset which I cluster with an HMM before the training in Rapidminer - The goal is to use the data from one state ("E1") to train my Deep Learning model and afterwards predict some Label ("TRUE"/"FALSE") in the rest of the dataset (so in principal every state, except E1)
• Plus, I have a class imbalance in my data, I have way more "FALSE" labels than "TRUE" ones

My way of implementing this problem in Rapidminer is this:

• I retrieve the two datasets and create weights for the different labels (TRUE gets assigned a weight of 10, FALSE gets assigned a weight of 1) for dealing with the class imbalance. Afterwards I sample 50% of the training set, and run it through a Deep Learning classifier including a Leave-One-Out Cross-Validation
• Afterwards I apply this model to the Test set and predict the Performance (Binomial, because it's a binomial label)

I appended the two input files (anonymized in a way, that I can post them here and they're still enough for training/testing) and my Process as an XML-file.

My question now is, if there are any pitfalls or any basic things I'm overlooking? I'm still quite a beginner in the Machine Learning department and a complete beginner in Rapidminer. I'm just not sure if my way is scientificly correct, or if it could be better implemented in Rapidminer.

Best regards and thanks for your help

Jakob

<有限公司ntext>








<参数键= " notification_email“价值= " / >




























<操作符= " true " class = " numerical_to_bin激活ominal" compatibility="7.5.003" expanded="true" height="82" name="Numerical to Binominal (2)" width="90" x="514" y="187">






















































<操作符= " true " class = " numerical_to_bin激活ominal" compatibility="7.5.003" expanded="true" height="82" name="Numerical to Binominal" width="90" x="581" y="34">























<有限公司nnect from_port="in 1" to_op="Numerical to Polynominal" to_port="example set input"/>
<有限公司nnect from_port="in 2" to_op="Generate Attributes (2)" to_port="example set input"/>
<有限公司nnect from_op="Generate Attributes (2)" from_port="example set output" to_op="Generate Attributes (5)" to_port="example set input"/>
<有限公司nnect from_op="Generate Attributes (5)" from_port="example set output" to_op="Set Role (3)" to_port="example set input"/>
<有限公司nnect from_op="Set Role (3)" from_port="example set output" to_op="Numerical to Binominal (2)" to_port="example set input"/>
<有限公司nnect from_op="Numerical to Binominal (2)" from_port="example set output" to_op="Sample (3)" to_port="example set input"/>
<有限公司nnect from_op="Sample (3)" from_port="example set output" to_port="out 2"/>
<有限公司nnect from_op="Numerical to Polynominal" from_port="example set output" to_op="Generate Attributes (3)" to_port="example set input"/>
<有限公司nnect from_op="Generate Attributes (3)" from_port="example set output" to_op="Generate Attributes (4)" to_port="example set input"/>
<有限公司nnect from_op="Generate Attributes (4)" from_port="example set output" to_op="Set Role (2)" to_port="example set input"/>
<有限公司nnect from_op="Set Role (2)" from_port="example set output" to_op="Numerical to Binominal" to_port="example set input"/>
<有限公司nnect from_op="Numerical to Binominal" from_port="example set output" to_op="Sample (2)" to_port="example set input"/>
<有限公司nnect from_op="Sample (2)" from_port="example set output" to_port="out 1"/>




































<参数键= " learning_rate " value = " 0.005 " / >


<参数键= "mentum_start" value="0.0"/>
<参数键= "mentum_ramp" value="1000000.0"/>
<参数键= "mentum_stable" value="0.0"/>
















<有限公司nnect from_port="training set" to_op="Deep Learning" to_port="training set"/>
<有限公司nnect from_op="Deep Learning" from_port="model" to_port="model"/>




































<有限公司nnect from_port="model" to_op="Apply Model" to_port="model"/>
<有限公司nnect from_port="test set" to_op="Apply Model" to_port="unlabelled data"/>
<有限公司nnect from_op="Apply Model" from_port="labelled data" to_op="Performance (2)" to_port="labelled data"/>
<有限公司nnect from_op="Performance (2)" from_port="performance" to_port="performance 1"/>
<有限公司nnect from_op="Performance (2)" from_port="example set" to_port="test set results"/>












<有限公司nnect from_op="Retrieve Vista_wo_E1" from_port="output" to_op="Set weight" to_port="in 2"/>
<有限公司nnect from_op="Retrieve Vista_E1" from_port="output" to_op="Set weight" to_port="in 1"/>
<有限公司nnect from_op="Set weight" from_port="out 1" to_op="Cross Validation" to_port="example set"/>
<有限公司nnect from_op="Set weight" from_port="out 2" to_op="Apply Model (2)" to_port="unlabelled data"/>
<有限公司nnect from_op="Cross Validation" from_port="model" to_op="Apply Model (2)" to_port="model"/>
<有限公司nnect from_op="Apply Model (2)" from_port="labelled data" to_port="result 1"/>

sgenzer

hello@jakob_roetner- I'm going to pass this on to others in the team who know the DL operators better than I. Maybe@jpuente?

jakob_roetner

Thank you very much for your effortHighly appreciated!