New answers tagged

5 votes
Accepted

Fixing outliers and normalizing a vector using R

Techniques of Exploratory Data Analysis (EDA) can help with this feature engineering problem. I want to emphasize how just a couple of well-chosen plots tell us, forcibly, how we ought to proceed. ...
user avatar
  • 294k
1 vote

Fixing outliers and normalizing a vector using R

If you want to force the vector of observations to have a distribution close to normal, you can use inverse normal scores transformation. There are a few different varieties such as Elfving, Blom, ...
user avatar
0 votes

How to distinguish numerical categorical (Ex: White = 1, Latino = 2 etc) from numerical continuous or discrete variables on a dataset?

I found this helpful A key distinction between ordinal categorical variables and discrete quantitative variables is that there is a uniform degree of difference within discrete quantitative variables. ...
user avatar
  • 31
0 votes

Linear Connected Layer before the DNN

In your proposed architecture, the weights to the second and to the third layer are arbitrary. If you divide the weight $w^1_i$ to the $i$-th node in the second layer (for the connection from the $i$...
user avatar
  • 6,303
0 votes

Optimal way to create a feature set?

If this fixed data can be similar or even partially equal for different time series and you expect the fixed data to contain information that is relevant for your time series analysis task (prediction?...
user avatar
  • 6,303
0 votes

Selecting features for classification model

The transformation exists to put tiny numbers on a more meaningful scale. Humans see tiny p-values like $0.00001627$ and $0.0003119$ as just "tiny numbers". Try taking $-\log_{10}$ of those ...
user avatar
  • 33.5k
4 votes

Improving a forest model by dropping features below a percent importance threshold?

In theory, you could take such an approach, but there's a number of caveats. Firstly, whether this is helpful for predictive performance will be rather case specific and it could certainly make your ...
user avatar
  • 23.4k

Top 50 recent answers are included