Please enable JavaScript.
Coggle requires JavaScript to display documents.
Success, risk and continuation (identify success criteria (start with…
Success, risk and continuation
identify success criteria
start with small, clear goals
is management on board with the project?
can the model drivers be visualized?
who will use the model?
how much value can the model produce?
support from subject matter experts and management makes the difference between success and failure in every project
management support important at end of project
model implementation into information workflow
value must be specified and communicated relative to costs and benefits
foresee risks
risks are difficult to identify
play devils advocate
some possible risks off the bat:
target leakage in the model
data may be missing or of insufficient quality
model is insufficiently predictive
four major types of risk
ethical risk
does your historical data contain unethical processes
loans not given based off race, ethnicity in past
always consider historic context of data in model
privacy concern
user internet history is sensitive
models can predict whether a FB user is homosexual
not ethical to use data in a way that is unnecessarily harmful to society members
cultural risk
culture of organization may inhibit AutoML from being successful
begin building/forming a "data-savvy" workplace culture
model risk
do you have access to relevant data
possibility for historical data to be subjective or outdated
environmental risk
organizations external environment
external env can change drastically - reducing success of model
be prepared to re-model in "black swan" situation
decide whether to continue
weigh risks against rewards
evaluate whether project should continue
possibility of running "pilot project" first
can also garner interest in more complete model/AutoML project