利用R语言做可重复性报告研究

插入R代码样例展示：

---

title: "R代码块"

output: html_document

---



### 不显示warnings信息

```{r,warning=FALSE}

a <- 1:4

b <- 1:3

a +b

```



***



#### 不显示代码

```{r,fig.align='left',echo=FALSE}

library(recharts)

g <- echartr(iris, Sepal.Width, Petal.Width, series =Species) %>%

  setSeries(symbolSize=8) %>%

  setSymbols(c('heart', 'arrow','diamond')) %>%

  setToolbox(show=FALSE)

g %>% addMarkPoint(series=unique(iris$Species),

                   data=data.frame(type="max",name="最大值"))

```

•Rmarkdown通常以html、word、pdf格式输出，但是也可实现slidy输出。只需将output设置为“slidy_presentation” 即可。

---

title: "Presentations with Slidy"

author: "Daniel xie"

date: "November 14, 2017"

output:

  slidy_presentation:

    incremental: yes

---



#### 使用 Markdown 的优点



-  专注你的文字内容而不是排版样式。

-  轻松的导出 HTML、PDF 和本身的 .md 文件。

-  纯文本内容，兼容所有的文本编辑器与字处理软件。

-  可读，直观。

-  适合所有人的写作语言。



#### RMarkdown简要语法



[RStudio](www.rstudio.com)



![](R.png)



**这是粗体**



*这是斜体*



<font color=#D87093>使用十六进制颜色值: color=#D87093 </font>



~~删除文本~~



#### 插入表格

```{r}

library(DT)

datatable(iris)

```



#### 输出模型结果

```{r}

fit <-lm(women)

summary(fit)

```



#### 打印图形

```{r,fig.align='center'}

par(mfrow=c(2,2))

plot(fit)

par(mfrow=c(1,1))

```



#### 交互可视化

```{r}

library(rpivotTable)

rpivotTable(mtcars)

```

实例：构建自动化报表

•Rmarkdown最大的优势是可以制作自动化报表，将人从繁琐的重复性工作中解放出来，更专注创造性的工作。

•头有这样的需求，针对大量的指标，需要对每个指标的时间序列进行研究，且需要区分不同的用户群，此时用rmarkdown来构建报表是最适宜的。

•模拟一份数据集，一共有10万个业务员，有10个过程指标（共60个字段），且业务员分为7种不同的类型：

•构建衡量时间序列的指标：波动性和幅度（利用差分和均值进行组合构建）

•然后利用衍生指标，分别对每个人群建立K均值聚类，查看不同组别的业务员在最近6个月的变化趋势，找出规律

•对于聚类结果，根据不同组别的簇中心，进行可视化展示。

---

title: "过程化指标聚类"

author: "谢佳标"

date: "2017年11月16日"

output: html_document

---



```{r,warning=FALSE,message=FALSE,echo=FALSE,pdi=300,fig.width=12}

set.seed(1234)

empno = seq(1,100000)

gu_flag = sample(c("C00","C12","C3","C45","L45","1A_under","1A"),100000,replace = T)

dat <- matrix(sample(c(1,2,3,50:500,1000,2000,5000),100000*60,replace = T),100000,60)

customer <- data.frame(empno,gu_flag,dat)

for(n in 1:10){

 customer_new <- customer[,c(1,2,(6*n-3):(6*n+2))] 

 # 计算差分

 customer_diff1 <- t(apply(customer_new[,c(-1,-2)],1,diff))

 # 增加实际值均值、（差分/上月）均值和差分为正的个数

 customer_data1 <- data.frame(empno = customer_new$empno,

                              gu_flag = customer_new$gu_flag,

                              mean = rowMeans(customer_new[,3:8],na.rm = T),

                              diff = rowMeans(customer_diff1/rowMeans(customer_new[,3:7]),na.rm = T),

                              positive = rowSums(ifelse(customer_diff1 > 0,1,0)))

 # 将diff变量中的-Inf与Inf、NA设置为0

 customer_data1$diff <- ifelse(is.infinite(customer_data1$diff),0,customer_data1$diff)

 customer_data1$diff <-  ifelse(is.na(customer_data1$diff),0,customer_data1$diff)

 # 按照gu_flag进行分组、分别建立K均值聚类模型

 gu_flag1 <- c("C00","C12","C3","C45","L45","1A_under","1A")

 for(i in 1:length(gu_flag1)){

   customer_data1_sub1 <- customer_data1[customer_data1$gu_flag==gu_flag1[i],]

   set.seed(1234)

   km_model <- kmeans(scale(customer_data1_sub1[,3:5]),3)

   # 给原始数据打上标签

   customer_new[customer_new$gu_flag==gu_flag1[i],'type'] <- km_model$cluster

 }

 # 按照gu_flag、type进行分组，分别统计最近六个月的平均值(按列求平均值)

 customer_sub_mean1 <- data.frame(aggregate(customer_new[,3:8],list(gu_flag=customer_new[,2],type=customer_new[,9]),

                                                                    FUN=function(x){mean(x,na.rm = T)}))

 colnames(customer_sub_mean1)[3:8] <- seq(6,1,-1)

 

# 展示数据集

 print(knitr::kable(customer_sub_mean1,digits = 2,padding=0))

 # 绘图

 library(reshape)

 md <- melt(customer_sub_mean1,id=c("gu_flag","type"))

 library(lattice)

 mygraph <- xyplot(value ~ variable|gu_flag,data=md,groups = type,type=c("l","p"),

                  pch=16,xlab=NULL,layout=c(7,1),

                  ylab=NULL,main=paste0("指标",n,"过程化指标聚类结果"))

 print(update(mygraph,

              panel = function(...){

                panel.grid(h=-1,v=-1)

                panel.xyplot(...,border="transparent")

              }))

}

```

效果如下：