Please enable JavaScript.
Coggle requires JavaScript to display documents.
Supervised ML - Coggle Diagram
Supervised ML
Linear Regression
Minimize error (loss) function
Modeling best practice
Use cost function to fit model
Develop multiple models
Compare model results and choose best
Types of errors
Workflow
[data with outcomes] + [model] --fit-->
[fitted model]
[data without outcomes] +
[fitted model]
--predict--> [predicted outcomes]
Intro
Model (equation)
Statistical
Good knowledge of underlying process
Machine learning
Poor knowledge of underlying process
Omits unimportant details (reduces complexity)
Small thing that captures larger thing
Retains important details
ML Framework
𝝮 : fit parameters
estimated from data
x: input
yp : output (predicted by model)
f
( ) : prediction function
J
(y, yp) : Loss
Error between y and yp
Train model to minimize
Estimates relationship between features and target
Hyperparameters
Used in ML workflow
Not components of model
Modeling approaches
Regression
Numerical y
Classification
Categorical y
Objective
Interpretation
Use 𝝮 to generate insights
Example: what's driving sales vs what sales will be
Prediction
Compare yp with y
Performance metrics
Risk having black-box model
