Logistic Regression, Prediction and ROC

Logistic Regression, Prediction and ROC - Log istic Reg r...

Info iconThis preview shows pages 1–4. Sign up to view the full content.

View Full Document Right Arrow Icon
2/17/2014 Logistic Regression, Prediction and ROC https://blackboard.uc.edu/bbcswebdav/pid-9566224-dt-content-rid-55868231_2/courses/14SS_BANA7046002/notes%284%29.html 1/15 Logistic Regression, Prediction and ROC The objective of this case is to get you understand logistic regression (binary classification) and some important ideas such as cross validation, ROC curve, cut-off probability. Code in this case is built upon lecture slides and Shaonan Tian's sample code. Input and sample data First load the credit scoring data. It is easy to load comma-separated values (CSV). credit.data <- read.csv("http://homepages.uc.edu/~maifg/7040/credit0.csv", header = T) Now split the data 90/10 as training/testing datasets: subset <- sample(nrow(credit.data), nrow(credit.data) * 0.9) credit.train = credit.data[subset, ] credit.test = credit.data[-subset, ] The training dataset has 63 variables, 4500 obs. colnames(credit.train)
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
2/17/2014 Logistic Regression, Prediction and ROC https://blackboard.uc.edu/bbcswebdav/pid-9566224-dt-content-rid-55868231_2/courses/14SS_BANA7046002/notes%284%29.html 2/15 ## [1] "id" "Y" "X2" "X3" "X4" "X5" "X6" ## [8] "X7" "X8" "X9" "X10_2" "X11_2" "X12_2" "X13_2" ## [15] "X14_2" "X15_2" "X15_3" "X15_4" "X15_5" "X15_6" "X16_2" ## [22] "X16_3" "X16_4" "X16_5" "X16_6" "X17_2" "X17_3" "X17_4" ## [29] "X17_5" "X17_6" "X18_2" "X18_3" "X18_4" "X18_5" "X18_6" ## [36] "X18_7" "X19_2" "X19_3" "X19_4" "X19_5" "X19_6" "X19_7" ## [43] "X19_8" "X19_9" "X19_10" "X20_2" "X20_3" "X20_4" "X21_2" ## [50] "X21_3" "X22_2" "X22_3" "X22_4" "X22_5" "X22_6" "X22_7" ## [57] "X22_8" "X22_9" "X22_10" "X22_11" "X23_2" "X23_3" "X24_2" Logistic Regression Let's build a logistic regression model based on all X variables. Note id is excluded from the model. credit.glm0 <- glm(Y ~ . - id, family = binomial, credit.train) You can view the result of the estimation: summary(credit.glm0) Note that there might be a warning message “ Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred ”. This happens because of a problem called quasi-complete separation. You can learn more here or here . The offending variable is X9 . We will choose to ignore the warning now. You can try fitting your model without X9 . The usual stepwise variable selection still works for logistic regression. caution: this will take a very long time . credit.glm.step <- step(credit.glm0) Or you can try model selection with BIC: credit.glm.step <- step(credit.glm0, k = log(nrow(credit.train)))
Background image of page 2
2/17/2014 Logistic Regression, Prediction and ROC https://blackboard.uc.edu/bbcswebdav/pid-9566224-dt-content-rid-55868231_2/courses/14SS_BANA7046002/notes%284%29.html 3/15 Prediction and Cross Validation Using Logistic Regression Now suppose there are 2 models we want to test, one with all X variables(credit.glm0), and one with X3, X8 and X11_2(credit.glm1). credit.glm1 <- glm(Y ~ X3 + X8 + X11_2, family = binomial, credit.train) AIC(credit.glm0) ## [1] 1713 AIC(credit.glm1) ## [1] 1891 BIC(credit.glm0) ## [1] 2110 BIC(credit.glm1) ## [1] 1916 Understanding classification decision making using logistic regression To get prediction from a logistic regression model, there are several steps you need to understand. Refer to textbook/slides for detailed math.
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Image of page 4
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

Page1 / 15

Logistic Regression, Prediction and ROC - Log istic Reg r...

This preview shows document pages 1 - 4. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online