Please enable JavaScript.
Coggle requires JavaScript to display documents.
Implement Document Maintain (Set up prediction system (Retraining model…
Implement Document Maintain
Set up prediction system
Retraining model
Environment being modelled changes subtly
Performance of a model was validated with data collection over same period
How to improve predictive ability
Click orange plus sign
100% of the data can be utilized to create the model
Rerun any model used as an input into blender at 100%
Estimated numbers so proceed with caution
Make model available in business process flow
Choose deployment model
User must determine what fits organization's needs
Drag and drop
accessed through the predict screen
up to five features
ID should be used
slow approach
No larger than 750MB
Application programming interface
Straight forward for those who can program R or Python
show example
introductory information on their use
top three reasons why person has been flagged
Data robot prime
Approximation of selected model
in-house responsibility
vendor independence
generate code
Batch
batch scoring is avaialble online
in place with spark
apache spark
widely distributed data processing environemnt
Document modeling process for reproductibility
Model documentation
Opportunity to do more desirable central machine learning work
Save time in the future
where did the data come from
how was it processed
what parameters were selected
Document where the data came from
electronic health record
document all steps taken in data robot
LogLoss optimization criteriod
autopilot to build models
process has been simplified
Create model monitoring and maintenence
potential problems
monitoring and maintenance plan
potential environment changes
business data environment or outside world
Extracted from ssytem may have changed EHR
strategies
fail rather than attempt to make the best of available data
column missing or column name altered
rerun model as soon as sufficent data is available
wait until target value is also available
regardless of data