Breaking News: Grepper is joining You.com. Read the official announcement!

model performance

Innocent Iguana answered on December 20, 2023 Popularity 8/10 Helpfulness 4/10

answer model performance

model performance

Comment

- Classification:
    - Accuracy is not a good metric for all cases. eg: class imbalanced dataset
    - accuracy = (TP + TN)/ (TP + TN + FP + FN)
    - precision = TP / (TP + FP)
    - High precision = lower false positive rate
    - recall = TP / (TP + FN)
    - High recall = lower false negative rate
    - f1 = 2 * (precision * recall) / (precision + recall)
    - ROC AUC = area under the curve of TP rate vs FP rate graph
    - jaccard score
    - log loss
- Regression:
    - r-squared : Percentage of variability of y explained by independent variable
    - RMSE : Root mean squared error. Average error in target
    - RSE : Residual standard error. Same as RMSE, except instead of length (n), it requires degree of freedom (n-1)
    - MAE : Average absolute error in target
    - VResidual vs fit plot : Fitted values on X axis vs Residual on Y axis. Good model has positive and negative values distributed evenly on both side of a line (gaussian noise).
    - Q-Q Plot : normal distribution on X axis vs dataset distribution on Y axis. Good model has linear relationship of equation line Y= mX + c
    - Distance location plot : Fitted values on X-axis vs root of standardized residuals on Y-axis. 
    - measurement of extreme values (outliers):
        - leverage : measurement of how extreme the explanatory variable values are
        - influence : how much the model would change if you leave the observation out of the dataset when modeling. (eg : cooks distance)
- Hyperparameter tuning = 
    - Hyperparameters are Parameters we specify before fitting the model
    - We compare model outcomes by changing these hyperparametes 
    - Use cross-validation to avoid overfitting
    - example : gridsearch
- Use boxplot of cross-validation results of different models to compare their distribution of scores

Popularity 8/10 Helpfulness 4/10 Language whatever

Source: Grepper

Tags: model performance whatever

Link to this answer
Share Copy Link

Contributed on Dec 20 2023

Innocent Iguana

0 Answers Avg Quality 2/10

model performance

Contents

More Related Answers

model performance

Grepper

Documentation

Social

Legal

Contact

Oops, You will need to install Grepper and log-in to perform this action.