Please enable JavaScript.
Coggle requires JavaScript to display documents.
Chapter 21-23 (Set up Prediction System (choose deployment strategy (drag…
Chapter 21-23
Set up Prediction System
retraining the model
the best model developed may be perform the best when practiced
choose deployment strategy
drag-and-drop
predict screen
slow approach to scoring
won't allow files bigger than 750MB
application programming interface (API)
straightforward if you know R or Python
allows access to reason codes, important for low-volume predictions
DataRobot Prime
creates an approximation of the selected model
not as accurate as original model, VERY comparable
Batch
use API to upload and score multiple large files in parallel
in-place with Spark
executable file to be used in an Apache Spark environment
Create Model Monitoring and Maintenance Plan
Potential Problems
potential environmental changes can be widely varied
sometimes hard to detect
Strategies
rerun the model as soon as sufficient new data is available
Document Modeling Process for Reproducibility
model documentation
documenting the model is where projects most often fail
by documenting the project, an analyst can save time on project