Please enable JavaScript.
Coggle requires JavaScript to display documents.
Ch. 2 Automated Machine Learning (8 Criteria for Auto ML excellence…
Ch. 2 Automated Machine Learning
important skill for data scientists
not coding
communicating in the language that all stakeholders understand
Machine Learning Life Cycle
Define Project Adjectives
Acquire & Explore Data
Model Data
Interpret & Communicate
Implement, Document & Maintain
not always linear process
might need to fix problems in previous stages
What is Auto ML?
automating complex tasks
require accuracy & speed beyond capabilities of humans
best learned without having to know complex working components
"under the hood" stuff
enables democratization of data science
available & understandable to most
AirBnB Case
Automated repetitive tasks
reduced model error by over 5%
important aspects
exploratory data analysis
feature engineering
algorithm selection and hyperparameter tuning
model diagnostics
What Auto ML is not
NOT automatic
still need decisions made by analyst
still need skillset to evaluate results
cannot yet solve
time series motifs
trajectory mining
Tools & platforms
context-specific tools
implemented in another system for a specific purpose
ex: Salesforce Einstein in Salesforce
general platforms
open source
ex: Python and R
commercial
ex: DataRobot
8 Criteria for Auto ML excellence
accuracy
productivity
ease of use
understanding and learning
resource availability
process transparency
generalization
recommended actions