Machine Learning

Linear regression

\hat Y = \hat\beta1 X + \hat\beta0

Linreg : \hat\beta_1

\frac{COV(X,Y)}{VAR(X)}=r\times\frac{\sigma_Y}{\sigma_X}

Linreg : \hat\beta_0

\mu_Y - \hat\beta_1\mu_X

MAE

\sum{i=1}^{n} \vert yi - \hat{y}_i \vert ^2

RSS = SSE

\sum{i=1}^{n} \big( \hat{y}i - \bar{y}_i \big)^2

Multilinreg : \hat Y

\sum\hat\beta_i X_i + \hat\beta_0

TSS = MSE = MSD

\sum^n{i=1}(yi-\bar y_i)^2

RSE

\sqrt \frac{SSE}{n-2}

R^2

1 - \frac{RSS}{TSS}

RMSE = RMSD

\sqrt{\frac{1}{n} \sum{i=1}^{n} \big( yi - \hat{y}_i \big)^2}

Binary logreg

\frac{\exp (\beta_0 + \sum\hat\beta_i Xi)}{1+ \exp (\beta_0 + \sum\hat\beta_i Xi)}

Classifier

KNN, LVQ, Naive Bayes, Decision Tree

Accuracy

TP + TN / nb predict

Recall ( 1 = min FN)

TP / TP+FN

FPRate

FP / TN+FP

specificity (TNRate)

1 - FPR

precision (1 = min FP)

TP / TP+FP

F-Measure

(2 x precision x recall) / (precision x recall)