1
$\begingroup$

This is a technical/conceptual question. I am not sure if this is the right place to ask. If not, please let me know, I will change it.

Question: I have some time series data with 12 room temperatures of a house. All are real number float32 values, ranging -5degC to 30degC.

I was thinking to create a prediction model for one of the room temperatures by using the other room temperatures as input. Now, when I calculate the correlation of the input features with desired output room temperature, using the Pearson method, they are usually very high. There are still 2-3 rooms, that are not very correlated.

Now, my understanding says, that if I use the features as input to the model whose correlation values are high, I will have better accuracy on a model, compared to the low correlated features.

Is it right? I am doing a prediction model to forecast future temperatures.

Let me know your thoughts!

$\endgroup$
0
$\begingroup$

Not necessarily. Correlation describes the linear relationship between 2 variables. As such, in the presence of other variables, it is possible for uncorrelated features to have significant explanatory power.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.