Hi All,
I’m trying to fit a neural network model using ore.neural(). There are around 150 input features. Almost thirty-three features are categorical variables such as client, project type, voltage type, tower type, to name a few.
For a neural network, I transformed each categorical variable into n-columns using one-hot encoding where n is the distinct values of that variable. However, there are some categorical variables like tower type that has approximately 1,700 distinct values. If I follow the same strategy, I’m likely to end up with a data frame comprising thousands of columns containing 1s and 0s.
Is this the recommended way of handling categorical variables using the ore.neural() in this use case?
Can I handle categorical variables differently using the ore.neural()?
Is there a way to automate this one-hot encoding transformation using the ore.neural()? Any example code or idea?
One can find the details of one-hot-encoding on the following link: https://machinelearningmastery.com/how-to-one-hot-encode-sequence-data-in-python/
Any help to get me achieve this issue efficiently will be greatly appreciated.
Many Thanks and
Kind Regards,
Bilal