Please enable JavaScript.
Coggle requires JavaScript to display documents.
REGRESSION - Coggle Diagram
REGRESSION
Evaluating performance
Prediction error = predicted value - actual value
predicted > actual = overprediction
Predicted < actual = underpredicted
Mean error
Positive = over
Negative = under
Mean absolute error
Absolute value of error
Magnitude of error in any direction
MAPE (percentage error)
relative to actual values
% indicates deviation of predictions
Total sum sqd error
square prediction errors
Root Mean Sqd Error
avg magnitude of error
STEPS
1) Model Construction - use training data
2) Model Testing - measure accuracy using test data
Training - Validation - Test (random split)
New Data - Prediction
Regression Trees
SSD at given node (not entropy)
Avg of outcome values = final node
Decision Tree - classify based on decision rules
Split rule = lowest possible value SSD (each data point)
0 = pure
KNN For Regression
Identify k nearest neighbors
Decide K - minimize RMSE
New prediction uses avg nearest neighbors values
can use weighted avg
Regression Tree Output
New observations tested against tree
discretize outcomes (segment)
Not all attributes must be included - like decision tree
PREDICTIVE
Regression - Continuous/numeric output
Classification - Categorical output
INPUT
Observations/Records
Attributes divided into 2
X - predictors
Y - outcome