"root_mean_squared_error using SVM"
Hi all,
I'm trying to generate some basic models/predictions using the set-up shown:
The THz data file contains just three columns: Time (integer), Density of people (integer), and Water vapour (integer).
The process works, in that I get the following:
However, in
Ah, ok, a different question than I guessed (I need to work on my predictive model :smileywink:)
Accuracy and precision would only be used for a classification problem but it looks like you try to predict numbers (called "regression"). You could turn this into a classification problem by applying some preprocessing step before (for example "Numerical to Polynominal"). Then you would get accuracy or precision instead.
If you want to see other performance measurements in general, you should try out different "Performance (...)" operators and change their parameters. They allow you to define exactly what type of performance measurement you want to calculate. The operator "Performance" just guesses what you might be interested in based on the type of the label column.
Cheers,
Ingo
Hi,
it's because you have a regression and not a classification task.
The accuracy is the relative number of correctly classified examples or in other words the percentage of correct predictions.
The precision is the relative number of correctly as positive classified examples among all examples classified as positive.
Greetings,
Sigrun
Brilliant - thanks very much.
It's the little things that trip you up - like the difference between regression and classification!
What both of you said makes perfect sense. I was asking the right question of the data and getting the appropriate performance metric for the answer, but I just didn't recognise it.
Thanks again.
Ah, ok, a different question than I guessed (I need to work on my predictive model :smileywink:)
Accuracy and precision would only be used for a classification problem but it looks like you try to predict numbers (called "regression"). You could turn this into a classification problem by applying some preprocessing step before (for example "Numerical to Polynominal"). Then you would get accuracy or precision instead.
If you want to see other performance measurements in general, you should try out different "Performance (...)" operators and change their parameters. They allow you to define exactly what type of performance measurement you want to calculate. The operator "Performance" just guesses what you might be interested in based on the type of the label column.
Cheers,
Ingo
Hi,
it's because you have a regression and not a classification task.
The accuracy is the relative number of correctly classified examples or in other words the percentage of correct predictions.
The precision is the relative number of correctly as positive classified examples among all examples classified as positive.
Greetings,
Sigrun
Hi,
Parts of your post got cut off - please post the rest :smileyhappy:
And just in case that your question is why all the predictions are so close together: Try a higher value for the parameter C for the SVM. Just a shot in the dark :smileywink:
Cheers,
Ingo