R语言做kaggle中California Housing Prices数据集

数据集来源

https://www.kaggle.com/c/house-prices-advanced-regression-techniques
数据集是kaggle里面的House Prices
加载所需要的包

library(tidyverse)
library(reshape2)

读取数据

housing = read.csv('housing.csv')

简单看一下数据

head(housing)

summary(housing)

• 可以看出total_bedrooms中有缺失值需要处理
• 再有就是ocean_proximity需要处理成哑变量
• total_bedrooms 和 total_rooms需要处理为mean_number_bedrooms and mean_number_rooms

看一下列名字

colnames(housing)

画图看一下各个变量的分布

ggplot(data = melt(housing), mapping = aes(x = value)) + 
    geom_histogram(bins = 30) + facet_wrap(~variable, scales = 'free_x')

数据预处理

中位数插值

housing$total_bedrooms[is.na(housing$total_bedrooms)] = median(housing$total_bedrooms , na.rm = TRUE)

将total_bedrooms 和 total_rooms处理为mean_number_bedrooms and mean_number_rooms

housing$mean_bedrooms = housing$total_bedrooms/housing$households
housing$mean_rooms = housing$total_rooms/housing$households

drops = c('total_bedrooms', 'total_rooms')

housing = housing[ , !(names(housing) %in% drops)]

head(housing)

下面是将ocean_proximity处理成哑变量

housing1 <- housing[c(-7,-8)]
housing1 <- scale(housing1)

library(fastDummies)
housing2 <- dummy_cols(housing,remove_first_dummy = F)

housing3 <- data.frame(housing2[11:15],housing1,housing$median_house_value)

library(randomForest)
library(caret)

将数据集的80%划分为训练集，20%划分为测试集

set.seed(1234)
#将数据集的80%划分为训练集，20%划分为测试集
trainIndex = createDataPartition(housing3$housing.median_house_value, p=0.8,
                                  list = FALSE,
                                  times = 1)
#createDatapartition会自动从y的各个level随机取出等比例的数据来，组成训练集
#训练集
dataTrain = housing3[trainIndex,]
#测试集
dataTest = housing3[-trainIndex,]

随机森林

library(randomForest)

names(dataTrain)

set.seed(1)

train_y = dataTrain[,'housing.median_house_value']
train_x = dataTrain[, names(dataTrain) !='housing.median_house_value']

head(train_y)
head(train_x)

rf_model = randomForest(train_x, y = train_y , ntree = 500, importance = TRUE)

names(rf_model)

变量重要性

rf_model$importance

包外误差

oob_prediction = predict(rf_model)

train_mse = mean(as.numeric((oob_prediction - train_y)^2))
oob_rmse = sqrt(train_mse)
oob_rmse

看看在模型测试集上的效果

test_y = dataTest[,'housing.median_house_value']
test_x = dataTest[, names(dataTest) !='housing.median_house_value']


y_pred = predict(rf_model , test_x)
test_mse = mean(((y_pred - test_y)^2))
test_rmse = sqrt(test_mse)
test_rmse