Neural network data optimization

Airbus_emitter

Good morning,

I am new to rapidminer and machine learning and I would like to know if there is a way in rapid miner to optimize the weight of some features used in a clustering.

I have a database with 2 million lines to train a neural network. I would like it to automatically know what weight should be given to each feature in order to obtain the number of clusters searched.

This number of clusters solution is also known in another database, so it would have to be used as a form of validation.

Is this possible, is there something like this already created?

Thank you very much in advance.

behnish

Hello, did you try Auto model for just one Model and then Open Process after completion with your data set? After Open Process you will find many hints about processes allowing to monitor the weight of features etc.

Telcontar120

If you know the cluster value for each record, then you can turn this into a supervised machine learning model, where the cluster is the label you are trying to predict. This makes optimization a bit more straightforward. With unsupervised ML algorithms, optimization can be a bit trickier since you need to define some performance metric with which to optimize the outcomes (and not all operators will be set up to support this).