Please enable JavaScript.
Coggle requires JavaScript to display documents.
Larsen Chapter 16 (Additional Data (Additional Features, Additional cases)…
Larsen Chapter 16
-
One-hot encoding
For any categorical features that fulfull certain requirements, a new feature is created for every category that exists within the original feature
oh lord theres so much going on and i honestly have no energy to retype what i just read. Basically REREAD THE CHAPTER HERE. but also really look at the specific parameters.
-
Accuracy Tradeoffs
Speed vs Accuracy
-
Eddicient Frontier
-
-
-
Always look here. Ex. if time is factor, effiicient frontier needs to be followed to the left to find the most efficient model
-
Speed
Do slowestg model first, if slowest model produces results more quickly than needed, ignore speed as criterion
-
If fast responses not needed, its okay if model can't keep up with peaks in prediction need
Blueprints
Cool fun fact, data robot imputes missing values on its own using (presumably) the average
Imputation
Click missing values imputed box in blueprint hen click on link to show documentation. This shows DataRobot code! How cool!
Basically this isn't helpful at all, but kinda cool because DAtaRobot employees use this to access model-specific code when needed
Standardization
Why standardize a numeric feature? BAsically some algorithms and some linear models dont like different standard deviation