I need to build a prediction model than contain all these methods SOM, DBSCAN, Linear Regression....

asiddiq

I need to build a prediction model than contain all these methods SOM, DBSCAN, Linear Regression, and LIME. I know there are some errors. I need to get a suggestion that has all my methods. (I have Dengue Fever Disease data + Weather Data). if anyone can draw a model to me it will be helpful and appreciate it.

LeMarc

Hi,

there is an operator called "Select Subprocess". Here you can include several algorithm.

To chose which Algorithm you want to use - change the parameter "select which" according to the number. So if you want to use a decistion tree the setting should be "select which - 3".

If you want several algorithms/steps to be done first before moving on to the next step, try using the operator "Subprocess".

rfuentealba

Hello,

Let's see.

• Retrieve your data.
• Convert your data from nominal to numerical if needed.
• SOM is Self-Organizing-Map, a dimensionality reduction algorithm. You should reduce your dimensions before clustering. (Like, why clustering if we still have some dimensions/attributes that we don't need?)
• DBSCAN is for clustering, so you should generate your clusters now. Clustering helps you obtaining a little more accuracy in scoring when using a sample, depending on where you put it.
• Now you should generate a sample to apply Linear Regression. This is the important part: you never use dimensionality reduction before the supervised algorithm. But you don't train your algorithm before LIME, you do it inside LIME. That is: put the LIME super operator, inside LIME you should use a Cross Validation or Split Validation to actually obtain the performance, and then inside your validation, you should put your Linear Regression on your training part and the classic Apply Model / Performance on the testing part.
  

Self-Organizing Map is part of RapidMiner, DBSCAN too. For LIME, you have to install the Operator Toolbox extension.

All the best,

Rod.