Create a R script or R markdown file.

# read the 'HousePrice.csv$\text{'},$ load it to a dataframe.

house $2$ read.csv$(\text{'}$HousePrice$.$csv$\text{'})$

# convert continuous house price values to labels

house$2$$Pricefactor$($with$($house$2,$ ifelse$(($house$2$$Price 'low','high'$)))$

# check number of samples in each category

table$($house$2$$Price$)$

# Divide as training and testing: $20\%$ test $80\%$ train and get the training

data size

sample$\_$size floor$(0\mathrm{.}8*$nrow$($house$))$

# check the training data size

sample$\_$size

# get train data index

train$\_$ind sample$($seq$\_$len$($nrow$($house$2)),$ size $=$ sample$\_$size$)$

# generate training and test dataset

train house$2[$train$\_$ind,$]$

test house$2[-$train$\_$ind,$]$

# use glm to build logistic model

glm$.$fit glm$($Price Sqft$\_$Area$+$Lot$\_$Area$+$Age$+$Crime, data $=$ train,

family $=$ binomial$)$

summary$($glm$.$fit$)$

data $=$ train$)$

Coefficients:

Null deviance: $3597\mathrm{.}9$ on $5835$ degrees of freedam Residual deviance: $2388\mathrm{.}4$ on $5831$ degrees of freedom $(70$ observatians deleted due to missingness$)$ AIC: $2318\mathrm{.}4$

# predict on test dataset

predictedprob predict$($glm$.$fit, newdata $=$ test, type $=$ "response"$)$

head$($predictedprob$)$

# check the probability

newdata data.frame$($test$Sqft$\_$Area, test$Price, predictedprob$)$

head$($newdata$)$

ggplot$($newdata$,$aes$($x$=$test.Sqft$\_$Area, y$=$predictedprob$))+$ geom$\_$point$()$

# evaluate the prediction results

glm$.$pred $=$ factor$($ifelse$($predictedprob $>0\mathrm{.}5,$ 'low','high'$))$

confusionMatrix$($test$Price, glm$.$pred$)$