数据分析实例--R语言如何对垃圾邮件进行分类

Structure of a Data Analysis

1  数据分析的步骤

l  Define the question

l  Define the ideal data set

l  Determine what data you can access

l  Obtain the data

l  Clean the data

l  Exploratory data analysis

l  Statistical prediction/model

l  Interpret results

l  Challenge results

l  Synthesize/write up results

l  Create reproducible code

2   A sample

1)    问题.

Can I automatically detect emails that are SPAM or not?

2)    具体化问题

Can I use quantitative characteristics of the emails to classify them as SPAM/HAM?

3)    获取数据

http://search.r-project.org/library/kernlab/html/spam.html

4)    取样

#if it isn't installed,please install the package first.

library(kernlab)

data(spam)

#perform the subsampling

set.seed(3435)

trainIndicator =rbinom(4601,size = 1,prob = 0.5)

table(trainIndicator)

trainSpam = spam[trainIndicator == 1, ]

testSpam = spam[trainIndicator == 0, ]

5)    初步分析

a)      Names：查看的列名

names(trainSpam)

b)      Head:查看前六行

head(trainSpam)

c)       Summaries：汇总

table(trainSpam$type)

d)      Plots:画图,查看垃圾邮件及非垃圾邮件的分布

plot(trainSpam$capitalAve ~ trainSpam$type)

上图分布不明显，我们取对数后，再看看

plot(log10(trainSpam$capitalAve + 1) ~ trainSpam$type)

e)      寻找预测的内在关系

plot(log10(trainSpam[, 1:4] + 1))

f)        试用层次聚类

hCluster = hclust(dist(t(trainSpam[, 1:57])))

plot(hCluster)

太乱了.不能发现些什么。老方法不是取log看看

hClusterUpdated = hclust(dist(t(log10(trainSpam[, 1:55] + 1))))

plot(hClusterUpdated)

6)    统计预测及建模

trainSpam$numType = as.numeric(trainSpam$type) - 1

costFunction = function(x, y) sum(x != (y > 0.5))

cvError = rep(NA, 55)

library(boot)

for (i in 1:55) {

lmFormula = reformulate(names(trainSpam)[i], response = "numType")

glmFit = glm(lmFormula, family = "binomial", data = trainSpam)

cvError[i] = cv.glm(trainSpam, glmFit, costFunction, 2)$delta[2]

}

## Which predictor has minimum cross-validated error?

names(trainSpam)[which.min(cvError)]

7)     检测

## Use the best model from the group

predictionModel = glm(numType ~ charDollar, family = "binomial", data = trainSpam)

## Get predictions on the test set

predictionTest = predict(predictionModel, testSpam)

predictedSpam = rep("nonspam", dim(testSpam)[1])

## Classify as `spam' for those with prob > 0.5

predictedSpam[predictionModel$fitted > 0.5] = "spam"

## Classification table 查看分类结果

table(predictedSpam, testSpam$type)

分类错误率：0.2243 =(61 + 458)/(1346 + 458 + 61 + 449)

8)    Interpret results（结果解释）

The fraction of charcters that are dollar signs can be used to predict if an email is Spam

Anything with more than 6.6% dollar signs is classified as Spam

More dollar signs always means more Spam under our prediction

Our test set error rate was 22.4%

9)    Challenge results

10)  Synthesize/write up results

11)   Create reproducible code