Dear fellow researchers,
We have a data set with 30 features, some the features in this set were derived from other features from the same set, and we are trying to discover such dependencies (note that we are not looking for correlation but actual Input -> output dependencies. Also note that, in our case, there can be multiple dependencies in the same feature set).
Can you suggest some technique to approach this problem.
Thanks in advance
You can actually try to use supervised methods using one variable as output and other as input.
If you achieve a good accuracy, then those variables are linked with each other.
Use also feature selection, it would give you the most relevant variables.
Yohann
Dear Shivangi Garde , here is an idea:
Train a Random Forest with all the features. Then look at the variable importance for each feature. The output column should have higher dependencies with the features that have higher importance value yielded from Random Forest Variable Importance.
@Sergey - Thanks for replying can you please elaborate a bit more on how PCA might be helpful in knowing which among the given features are Dependent and Independent variables.
@ Abdullah Al Imran - Thank you for replying, but isn't it the case that for training random forest I would first need a target. In our case we don't know which one is target and there can be multiple on them too.
Principal component analysis generates a new set of variables, called principal components. Each principal component is a linear combination of the original variables. All the principal components are orthogonal to each other, so there is no redundant information. The principal components as a whole form an orthogonal basis for the space of the data.
- Badges
- Science topic