Data Science with R Decision Trees

相似文件 换一批

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

Data Science with R Decision Trees

Decision trees are widely used in data mining and well supported in R (R Core Team, 2014). Decision tree learning deploys a divide and conquer approach, known as recursive partitioning. It is usually implemented as a greedy search using information gain or the Gini index to select the best input variable on which to partition our dataset at each step.
This Module introduces rattle (Williams, 2014) and rpart (Therneau and Atkinson, 2014) for building decision trees. We begin with a step-by-step example of building a decision tree using Rattle, and then illustrate the process using R begining with Section 14. We cover both classification trees and regression trees.
The required packages for this module include:

• library(rattle) # GUI for building trees and fancy tree plot
• library(rpart) # Popular decision tree algorithm
• library(rpart.plot) # Enhanced tree plots
• library(party) # Alternative decision tree algorithm
• library(partykit) # Convert rpart object to BinaryTree
• library(RWeka) # Weka decision tree J48.
• library(C50) # Original C5.0 implementation.
  

As we work through this chapter, new R commands will be introduced. Be sure to review the command’s documentation and understand what the command does. You can ask for help using the ? command as in:
?read.csv
We can obtain documentation on a particular package using the help= option of library():
library(help=rattle)
This chapter is intended to be hands on. To learn effectively, you are encouraged to have R running (e.g., RStudio) and to run all the commands as they appear here. Check that you get the same output, and you understand the output. Try some variations. Explore.

本帖隐藏的内容

Data Science with R Decision Trees.rar (1.14 MB) 本附件包括：

• Data Science with R Decision Trees.pdf

2015-6-13 08:07:04 上传

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏2 回帖

关键词：Data Science Decision Science Trees rees learning usually search 2014

1. ## ----module, echo=FALSE, results="asis"----------------------------------
   
2. Module <- "DTreesO"
   
3. cat(paste0("\\newcommand{\\Module}{", Module, "}"))
   
4. 
   
5. 
   
6. ## ----setup, child="mycourse.Rnw"-----------------------------------------
   
7. 
   
8. ## ----setup_options, include=FALSE----------------------------------------
   
9. library(knitr)
   
10. library(xtable)
    
11. 
    
12. opts_chunk$set(cache=FALSE)
    
13. 
    
14. opts_chunk$set(out.width='0.8\\textwidth')
    
15. opts_chunk$set(fig.align='center')
    
16. 
    
17. opts_chunk$set(src.top=NULL)
    
18. opts_chunk$set(src.bot=NULL)
    
19. opts_chunk$set(out.lines=4)
    
20. opts_chunk$set(out.truncate=80)
    
21. 
    
22. opts_chunk$set(fig.path=sprintf("figures/%s/", Module))
    
23. opts_chunk$set(cache.path=sprintf("cache/%s/", Module))
    
24. opts_chunk$set(bib.file=paste0(Module, ".bib"))
    
25. 
    
26. # Leave code as I have formatted it.
    
27. 
    
28. opts_chunk$set(tidy=FALSE)
    
29. 
    
30. # Hooks
    
31. 
    
32. # Allow auto crop of base graphics plots when crop=TRUE.
    
33. 
    
34. knit_hooks$set(crop=hook_pdfcrop)
    
35. 
    
36. # Truncate long lines and long output
    
37. 
    
38. hook_output <- knit_hooks$get("output")
    
39. hook_source <- knit_hooks$get("source")
    
40. knit_hooks$set(output=function(x, options)
    
41. {
    
42.   if (options$results != "asis")
    
43.   {
    
44.     # Split string into separate lines.
    
45.     x <- unlist(stringr::str_split(x, "\n"))
    
46.     # Trim to the number of lines specified.
    
47.     if (!is.null(n <- options$out.lines))
    
48.     {
    
49.       if (length(x) > n)
    
50.       {
    
51.         # Truncate the output.
    
52.         x <- c(head(x, n), "....\n")
    
53.       }
    
54.     }
    
55.     # Truncate each line to length specified.
    
56.     if (!is.null(m <- options$out.truncate))
    
57.     {
    
58.       len <- nchar(x)
    
59.       x[len>m] <- paste0(substr(x[len>m], 0, m-3), "...")
    
60.     }
    
61.     # Paste lines back together.
    
62.     x <- paste(x, collapse="\n")
    
63.     # Replace ' = ' with '=' - my preference. Hopefully won't
    
64.     # affect things inappropriately.
    
65.     x <- gsub(" = ", "=", x)
    
66.   }
    
67.   hook_output(x, options)
    
68. },
    
69. source=function(x, options)
    
70. {
    
71.   # Split string into separate lines.
    
72.   x <- unlist(stringr::str_split(x, "\n"))
    
73.   # Trim to the number of lines specified.
    
74.   if (!is.null(n <- options$src.top))
    
75.   {
    
76.     if (length(x) > n)
    
77.     {
    
78.       # Truncate the output.
    
79.       if (is.null(m <-options$src.bot)) m <- 0
    
80.       x <- c(head(x, n+1), "\n....\n", tail(x, m+2))
    
81.    }
    
82.   }
    
83.   # Paste lines back together.
    
84.   x <- paste(x, collapse="\n")
    
85.   hook_source(x, options)
    
86. })
    
87. 
    
88. # Optionally allow R Code chunks to be environments so we can refer to them.
    
89. 
    
90. knit_hooks$set(rcode=function(before, options, envir)
    
91. {
    
92.   if (before)
    
93.     sprintf('\\begin{rcode}\\label{%s}\\hfill{}', options$label)
    
94.   else
    
95.     '\\end{rcode}'
    
96. })
    
97. 
    
98. 
    
99. 
    
100. ## ----load_packages, message=FALSE----------------------------------------
     
101. library(rattle)         # GUI for building trees and fancy tree plot
     
102. library(rpart)          # Popular decision tree algorithm
     
103. library(rpart.plot)     # Enhanced tree plots
     
104. library(party)          # Alternative decision tree algorithm
     
105. library(partykit)       # Convert rpart object to BinaryTree
     
106. library(RWeka)          # Weka decision tree J48.
     
107. library(C50)            # Original C5.0 implementation.
     
108. 
     
109. 
     
110. ## ----echo=FALSE----------------------------------------------------------
     
111. library(RColorBrewer)
     
112. 
     
113. 
     
114. ## ----documentation, child="documentation.Rnw", eval=TRUE-----------------
     
115. 
     
116. 
     
117. ## ----help_library, eval=FALSE, tidy=FALSE--------------------------------
     
118. ## ?read.csv
     
119. 
     
120. 
     
121. ## ----help_package, eval=FALSE--------------------------------------------
     
122. ## library(help=rattle)
     
123. 
     
124. 
     
125. ## ----record_start_time, echo=FALSE---------------------------------------
     
126. start.time <- proc.time()
     
127. 
     
128. 
     
129. ## ----generate_bib, echo=FALSE, message=FALSE, warning=FALSE--------------
     
130. # Write all packages in the current session to a bib file
     
131. if (is.null(opts_chunk$get("bib.file"))) opts_chunk$set(bib.file="Course.bib")
     
132. write_bib(sub("^.*/", "", grep("^/", searchpaths(), value=TRUE)),
     
133.           file=opts_chunk$get("bib.file"))
     
134. system(paste("cat extra.bib >>", opts_chunk$get("bib.file")))
     
135. # Fix up specific issues.
     
136. # R-earth
     
137. system(paste("perl -pi -e 's|. Derived from .*$|},|'",
     
138.              opts_chunk$get("bib.file")))
     
139. # R-randomForest
     
140. system(paste("perl -pi -e 's|Fortran original by Leo Breiman",
     
141.              "and Adele Cutler and R port by|Leo Breiman and",
     
142.              "Adele Cutler and|'", opts_chunk$get("bib.file")))
     
143. # R-C50
     
144. system(paste("perl -pi -e 's|. C code for C5.0 by R. Quinlan|",
     
145.              " and J. Ross Quinlan|'", opts_chunk$get("bib.file")))
     
146. # R-caret
     
147. system(paste("perl -pi -e 's|. Contributions from|",
     
148.              " and|'", opts_chunk$get("bib.file")))
     
149. # Me
     
150. system(paste("perl -pi -e 's|Graham Williams|",
     
151.              "Graham J Williams|'", opts_chunk$get("bib.file")))
     
152. 
     
153. 
     
154. 
     
155. 
     
156. ## ----start_rattle, eval=FALSE--------------------------------------------
     
157. ## library(rattle)
     
158. ## rattle()
     
159. 
     
160. 
     
161. ## ----rattle_commands, eval=FALSE-----------------------------------------
     
162. ## set.seed(42)
     
163. ## library(rattle)
     
164. ## library(rpart)
     
165. ## ds     <- weather
     
166. ## target <- "RainTomorrow"
     
167. ## nobs   <- nrow(ds)
     
168. ## form   <- formula(paste(target, "~ ."))
     
169. ## train  <- sample(nobs, 0.70 * nobs)
     
170. ## vars   <- -c(1,2,23)
     
171. ## model  <- rpart(form, ds[train, vars], parms=list(split="information"))
     
172. 
     
173. 
     
174. ## ----prepare_data_weather------------------------------------------------
     
175. set.seed(1426)
     
176. library(rattle)
     
177. data(weather)
     
178. dsname       <- "weather"
     
179. ds           <- get(dsname)
     
180. id           <- c("Date", "Location")
     
181. target       <- "RainTomorrow"
     
182. risk         <- "RISK_MM"
     
183. ignore       <- c(id, if (exists("risk")) risk)
     
184. (vars        <- setdiff(names(ds), ignore))
     
185. inputs       <- setdiff(vars, target)
     
186. (nobs        <- nrow(ds))
     
187. (numerics    <- intersect(inputs, names(ds)[which(sapply(ds[vars], is.numeric))]))
     
188. (categorics  <- intersect(inputs, names(ds)[which(sapply(ds[vars], is.factor))]))
     
189. (form        <- formula(paste(target, "~ .")))
     
190. length(train <- sample(nobs, 0.7*nobs))
     
191. length(test  <- setdiff(seq_len(nobs), train))
     
192. actual       <- ds[test, target]
     
193. risks        <- ds[test, risk]
     
194. 
     
195. 
     
196. ## ----check_dataset, out.lines=5------------------------------------------
     
197. dim(ds)
     
198. names(ds)
     
199. head(ds)
     
200. tail(ds)
     
201. str(ds)
     
202. summary(ds)
     
203. 
     
204. 
     
205. ## ----dt_model------------------------------------------------------------
     
206. model <- rpart(formula=form, data=ds[train, vars])
     
207. 
     
208. 
     
209. ## ----dt_model_no_argument_names------------------------------------------
     
210. str(rpart)
     
211. model <- rpart(form, ds[train, vars])
     
212. 
     
213. 
     
214. ## ----display_model, out.lines=NULL---------------------------------------
     
215. model
     
216. 
     
217. 
     
218. ## ----dt_model_summary, out.lines=40--------------------------------------
     
219. summary(model)
     
220. 
     
221. 
     
222. ## ----dt_model_printcp, out.lines=NULL------------------------------------
     
223. printcp(model)
     
224. 
     
225. 
     
226. ## ----dt_model_plotcp,----------------------------------------------------
     
227. plotcp(model)
     
228. 
     
229. 
     
230. ## ----tmodel_weatherAUS_for_cp--------------------------------------------
     
231. tmodel <- rpart(form, weatherAUS[vars])
     
232. plotcp(tmodel)
     
233. 
     
234. 
     
235. ## ----tmodel_weatherAUS_cp_0----------------------------------------------
     
236. tmodel <- rpart(form, weatherAUS[vars], control=rpart.control(cp=0))
     
237. plotcp(tmodel)
     
238. 
     
239. 
     
240. ## ----out.lines=NULL------------------------------------------------------
     
241. tmodel$cptable[c(1:5,22:29, 80:83),]
     
242. 
     
243. 
     
244. ## ------------------------------------------------------------------------
     
245. model$variable.importance
     
246. 
     
247. 
     
248. ## ----message=FALSE-------------------------------------------------------
     
249. predicted <- predict(model, ds[test, vars], type="prob")[,2]
     
250. riskchart(predicted, actual, risks)
     
251. 
     
252. 
     
253. ## ------------------------------------------------------------------------
     
254. predicted <- predict(model, ds[test, vars], type="class")
     
255. sum(actual != predicted)/length(predicted) # Overall error rate
     
256. round(100*table(actual, predicted, dnn=c("Actual", "Predicted"))/length(predicted))
     
257. 
     
258. 
     
259. ## ------------------------------------------------------------------------
     
260. asRules.rpart <- function(model)
     
261. {
     
262.   if (!inherits(model, "rpart")) stop("Not a legitimate rpart tree")
     
263.   #
     
264.   # Get some information.
     
265.   #
     
266.   frm     <- model$frame
     
267.   names   <- row.names(frm)
     
268.   ylevels <- attr(model, "ylevels")
     
269.   ds.size <- model$frame[1,]$n
     
270.   #
     
271.   # Print each leaf node as a rule.
     
272.   #
     
273.   for (i in 1:nrow(frm))
     
274.   {
     
275.     if (frm[i,1] == "<leaf>")
     
276.     {
     
277.       # The following [,5] is hardwired - needs work!
     
278.       cat("\n")
     
279.       cat(sprintf(" Rule number: %s ", names[i]))
     
280.       cat(sprintf("[yval=%s cover=%d (%.0f%%) prob=%0.2f]\n",
     
281.                   ylevels[frm[i,]$yval], frm[i,]$n,
     
282.                   round(100*frm[i,]$n/ds.size), frm[i,]$yval2[,5]))
     
283.       pth <- path.rpart(model, nodes=as.numeric(names[i]), print.it=FALSE)
     
284.       cat(sprintf("   %s\n", unlist(pth)[-1]), sep="")
     
285.     }
     
286.   }
     
287. }

复制代码

1. ## ----out.lines=14--------------------------------------------------------
   
2. asRules(model)
   
3. 
   
4. 
   
5. ## ----basic_plot, echo=FALSE----------------------------------------------
   
6. plot(model)
   
7. text(model)
   
8. 
   
9. 
   
10. ## ----basic_plot, eval=FALSE----------------------------------------------
    
11. ## plot(model)
    
12. ## text(model)
    
13. 
    
14. 
    
15. ## ----basic_plot_uniform, echo=FALSE--------------------------------------
    
16. plot(model, uniform=TRUE)
    
17. text(model)
    
18. 
    
19. 
    
20. ## ----basic_plot_uniform, eval=FALSE--------------------------------------
    
21. ## plot(model, uniform=TRUE)
    
22. ## text(model)
    
23. 
    
24. 
    
25. ## ----basic_plot_extra, echo=FALSE----------------------------------------
    
26. plot(model, uniform=TRUE)
    
27. text(model, use.n=TRUE, all=TRUE, cex=.8)
    
28. 
    
29. 
    
30. ## ----basic_plot_extra, eval=FALSE----------------------------------------
    
31. ## plot(model, uniform=TRUE)
    
32. ## text(model, use.n=TRUE, all=TRUE, cex=.8)
    
33. 
    
34. 
    
35. ## ----fancy_plot, message=FALSE, echo=FALSE-------------------------------
    
36. fancyRpartPlot(model)
    
37. 
    
38. 
    
39. ## ----fancy_plot, message=FALSE, eval=FALSE-------------------------------
    
40. ## fancyRpartPlot(model)
    
41. 
    
42. 
    
43. ## ----prp_default, echo=FALSE---------------------------------------------
    
44. prp(model)
    
45. 
    
46. 
    
47. ## ----prp_default, eval=FALSE---------------------------------------------
    
48. ## prp(model)
    
49. 
    
50. 
    
51. ## ----prp_fav, echo=FALSE-------------------------------------------------
    
52. prp(model, type=2, extra=104, nn=TRUE, fallen.leaves=TRUE,
    
53.     faclen=0, varlen=0, shadow.col="grey", branch.lty=3)
    
54. 
    
55. 
    
56. ## ----prp_fav, eval=FALSE-------------------------------------------------
    
57. ## prp(model, type=2, extra=104, nn=TRUE, fallen.leaves=TRUE,
    
58. ##     faclen=0, varlen=0, shadow.col="grey", branch.lty=3)
    
59. 
    
60. 
    
61. ## ----prp_colour, echo=FALSE----------------------------------------------
    
62. col <- c("#FD8D3C", "#FD8D3C", "#FD8D3C", "#BCBDDC",
    
63.          "#FDD0A2", "#FD8D3C", "#BCBDDC")
    
64. prp(model, type=2, extra=104, nn=TRUE, fallen.leaves=TRUE,
    
65.     faclen=0, varlen=0, shadow.col="grey", branch.lty=3, box.col=col)
    
66. 
    
67. 
    
68. ## ----prp_colour, eval=FALSE----------------------------------------------
    
69. ## col <- c("#FD8D3C", "#FD8D3C", "#FD8D3C", "#BCBDDC",
    
70. ##          "#FDD0A2", "#FD8D3C", "#BCBDDC")
    
71. ## prp(model, type=2, extra=104, nn=TRUE, fallen.leaves=TRUE,
    
72. ##     faclen=0, varlen=0, shadow.col="grey", branch.lty=3, box.col=col)
    
73. 
    
74. 
    
75. ## ----prp_label_nodes, echo=FALSE-----------------------------------------
    
76. prp(model, type=1)
    
77. 
    
78. 
    
79. ## ----prp_label_nodes, eval=FALSE-----------------------------------------
    
80. ## prp(model, type=1)
    
81. 
    
82. 
    
83. ## ----prp_label_below, echo=FALSE-----------------------------------------
    
84. prp(model, type=2)
    
85. 
    
86. 
    
87. ## ----prp_label_below, eval=FALSE-----------------------------------------
    
88. ## prp(model, type=2)
    
89. 
    
90. 
    
91. ## ----prp_split_labels, echo=FALSE----------------------------------------
    
92. prp(model, type=3)
    
93. 
    
94. 
    
95. ## ----prp_split_labels, eval=FALSE----------------------------------------
    
96. ## prp(model, type=3)
    
97. 
    
98. 
    
99. ## ----prp_interior_lables, echo=FALSE-------------------------------------
    
100. prp(model, type=4)
     
101. 
     
102. 
     
103. ## ----prp_interior_lables, eval=FALSE-------------------------------------
     
104. ## prp(model, type=4)
     
105. 
     
106. 
     
107. ## ----prp_num_obs, echo=FALSE---------------------------------------------
     
108. prp(model, type=2, extra=1)
     
109. 
     
110. 
     
111. ## ----prp_num_obs, eval=FALSE---------------------------------------------
     
112. ## prp(model, type=2, extra=1)
     
113. 
     
114. 
     
115. ## ----prp_per_obs, echo=FALSE---------------------------------------------
     
116. prp(model, type=2, extra=101)
     
117. 
     
118. 
     
119. ## ----prp_per_obs, eval=FALSE---------------------------------------------
     
120. ## prp(model, type=2, extra=101)
     
121. 
     
122. 
     
123. ## ----prp_class_rate, echo=FALSE------------------------------------------
     
124. prp(model, type=2, extra=2)
     
125. 
     
126. 
     
127. ## ----prp_class_rate, eval=FALSE------------------------------------------
     
128. ## prp(model, type=2, extra=2)
     
129. 
     
130. 
     
131. ## ----prp_add_per_obs, echo=FALSE-----------------------------------------
     
132. prp(model, type=2, extra=102)
     
133. 
     
134. 
     
135. ## ----prp_add_per_obs, eval=FALSE-----------------------------------------
     
136. ## prp(model, type=2, extra=102)
     
137. 
     
138. 
     
139. ## ----prp_miss_rate, echo=FALSE-------------------------------------------
     
140. prp(model, type=2, extra=3)
     
141. 
     
142. 
     
143. ## ----prp_miss_rate, eval=FALSE-------------------------------------------
     
144. ## prp(model, type=2, extra=3)
     
145. 
     
146. 
     
147. ## ----prp_prob_class, echo=FALSE------------------------------------------
     
148. prp(model, type=2, extra=4)
     
149. 
     
150. 
     
151. ## ----prp_prob_class, eval=FALSE------------------------------------------
     
152. ## prp(model, type=2, extra=4)
     
153. 
     
154. 
     
155. ## ----prp_prob_class_per_obs, echo=FALSE----------------------------------
     
156. prp(model, type=2, extra=104)
     
157. 
     
158. 
     
159. ## ----prp_prob_class_per_obs, eval=FALSE----------------------------------
     
160. ## prp(model, type=2, extra=104)
     
161. 
     
162. 
     
163. ## ----prp_only_prob, echo=FALSE-------------------------------------------
     
164. prp(model, type=2, extra=5)
     
165. 
     
166. 
     
167. ## ----prp_only_prob, eval=FALSE-------------------------------------------
     
168. ## prp(model, type=2, extra=5)
     
169. 
     
170. 
     
171. ## ----prp_second_class, echo=FALSE----------------------------------------
     
172. prp(model, type=2, extra=6)
     
173. 
     
174. 
     
175. ## ----prp_second_class, eval=FALSE----------------------------------------
     
176. ## prp(model, type=2, extra=6)
     
177. 
     
178. 
     
179. ## ----prp_second_class_per_obs, echo=FALSE--------------------------------
     
180. prp(model, type=2, extra=106)
     
181. 
     
182. 
     
183. ## ----prp_second_class_per_obs, eval=FALSE--------------------------------
     
184. ## prp(model, type=2, extra=106)
     
185. 
     
186. 
     
187. ## ----prp_second_class_only_prob, echo=FALSE------------------------------
     
188. prp(model, type=2, extra=7)
     
189. 
     
190. 
     
191. ## ----prp_second_class_only_prob, eval=FALSE------------------------------
     
192. ## prp(model, type=2, extra=7)
     
193. 
     
194. 
     
195. ## ----prp_extra_8, echo=FALSE---------------------------------------------
     
196. prp(model, type=2, extra=8)
     
197. 
     
198. 
     
199. ## ----prp_extra_8, eval=FALSE---------------------------------------------
     
200. ## prp(model, type=2, extra=8)
     
201. 
     
202. 
     
203. ## ----prp_extra_9, echo=FALSE---------------------------------------------
     
204. prp(model, type=2, extra=9)
     
205. 
     
206. 
     
207. ## ----prp_extra_9, eval=FALSE---------------------------------------------
     
208. ## prp(model, type=2, extra=9)
     
209. 
     
210. 
     
211. ## ----prp_extra_100, echo=FALSE-------------------------------------------
     
212. prp(model, type=2, extra=100)
     
213. 
     
214. 
     
215. ## ----prp_extra_100, eval=FALSE-------------------------------------------
     
216. ## prp(model, type=2, extra=100)
     
217. 
     
218. 
     
219. ## ----prp_extra_106, echo=FALSE-------------------------------------------
     
220. prp(model, type=2, extra=106, nn=TRUE)
     
221. 
     
222. 
     
223. ## ----prp_extra_106, eval=FALSE-------------------------------------------
     
224. ## prp(model, type=2, extra=106, nn=TRUE)
     
225. 
     
226. 
     
227. ## ----prp_extra_106_ni, echo=FALSE----------------------------------------
     
228. prp(model, type=2, extra=106, nn=TRUE, ni=TRUE)
     
229. 
     
230. 
     
231. ## ----prp_extra_106_ni, eval=FALSE----------------------------------------
     
232. ## prp(model, type=2, extra=106, nn=TRUE, ni=TRUE)
     
233. 
     
234. 
     
235. ## ----prp_extra_106_fallen, echo=FALSE------------------------------------
     
236. prp(model, type=2, extra=106, nn=TRUE, fallen.leaves=TRUE)
     
237. 
     
238. 
     
239. ## ----prp_extra_106_fallen, eval=FALSE------------------------------------
     
240. ## prp(model, type=2, extra=106, nn=TRUE, fallen.leaves=TRUE)
     
241. 
     
242. 
     
243. ## ----prp_extra_106_fallen_branch, echo=FALSE-----------------------------
     
244. prp(model, type=2, extra=106, nn=TRUE, fallen.leaves=TRUE,
     
245.     branch=0.5)
     
246. 
     
247. 
     
248. ## ----prp_extra_106_fallen_branch, eval=FALSE-----------------------------
     
249. ## prp(model, type=2, extra=106, nn=TRUE, fallen.leaves=TRUE,
     
250. ##     branch=0.5)
     
251. 
     
252. 
     
253. ## ----prp_extra_106_faclen, echo=FALSE------------------------------------
     
254. prp(model, type=2, extra=106, nn=TRUE, fallen.leaves=TRUE,
     
255.     faclen=0)
     
256. 
     
257. 
     
258. ## ----prp_extra_106_faclen, eval=FALSE------------------------------------
     
259. ## prp(model, type=2, extra=106, nn=TRUE, fallen.leaves=TRUE,
     
260. ##     faclen=0)
     
261. 
     
262. 
     
263. ## ----prp_axtra_106_shadow, echo=FALSE------------------------------------
     
264. prp(model, type=2, extra=106, nn=TRUE, fallen.leaves=TRUE,
     
265.     shadow.col="grey")
     
266. 
     
267. 
     
268. ## ----prp_axtra_106_shadow, eval=FALSE------------------------------------
     
269. ## prp(model, type=2, extra=106, nn=TRUE, fallen.leaves=TRUE,
     
270. ##     shadow.col="grey")
     
271. 
     
272. 
     
273. ## ----prp_extra_106_branch, echo=FALSE------------------------------------
     
274. prp(model, type=2, extra=106, nn=TRUE, fallen.leaves=TRUE,
     
275.     branch.lty=3)
     
276. 
     
277. 
     
278. ## ----prp_extra_106_branch, eval=FALSE------------------------------------
     
279. ## prp(model, type=2, extra=106, nn=TRUE, fallen.leaves=TRUE,
     
280. ##     branch.lty=3)
     
281. 
     
282. 
     
283. ## ----eval=FALSE----------------------------------------------------------
     
284. ## plot(c(0,1), c(0,0), type="l", axes=FALSE, xlab=NA, ylab=NA, lty=2)
     
285. ## plot(c(0,1), c(0,0), type="l", axes=FALSE, xlab=NA, ylab=NA, lty="dashed")
     
286. ## plot(c(0,1), c(0,0), type="l", axes=FALSE, xlab=NA, ylab=NA, lty="44")
     
287. 
     
288. 
     
289. ## ----eval=FALSE----------------------------------------------------------
     
290. ## install.packages("partykit", repos="http://R-Forge.R-project.org")
     
291. ## library(partykit)
     
292. 
     
293. 
     
294. ## ----fig.width=14, out.width="\\textwidth"-------------------------------
     
295. class(model)
     
296. plot(as.party(model))
     
297. 
     
298. 
     
299. ## ----out.lines=15--------------------------------------------------------
     
300. print(as.party(model))
     
301. 
     
302. 
     
303. ## ----message=FALSE-------------------------------------------------------
     
304. library(partykit)
     
305. model <- ctree(formula=form, data=ds[train, vars])
     
306. 
     
307. 
     
308. ## ----out.lines=NULL------------------------------------------------------
     
309. model
     
310. 
     
311. 
     
312. ## ----message=FALSE-------------------------------------------------------
     
313. predicted <- predict(model, ds[test, vars], type="prob")[,2]
     
314. riskchart(predicted, actual, risks)
     
315. 
     
316. 
     
317. ## ------------------------------------------------------------------------
     
318. predicted <- predict(model, ds[test, vars], type="response")
     
319. sum(actual != predicted)/length(predicted) # Overall error rate
     
320. round(100*table(actual, predicted, dnn=c("Actual", "Predicted"))/length(predicted))
     
321. 
     
322. 
     
323. ## ------------------------------------------------------------------------
     
324. plot(model)
     
325. 
     
326. 
     
327. ## ------------------------------------------------------------------------
     
328. library(RWeka)
     
329. model <- J48(formula=form, data=ds[train, vars])
     
330. 
     
331. 
     
332. ## ----out.lines=NULL------------------------------------------------------
     
333. model
     
334. 
     
335. 
     
336. ## ----message=FALSE-------------------------------------------------------
     
337. predicted <- predict(model, ds[test, vars], type="prob")[,2]
     
338. riskchart(predicted, actual, risks)
     
339. 
     
340. 
     
341. ## ------------------------------------------------------------------------
     
342. predicted <- predict(model, ds[test, vars], type="class")
     
343. sum(actual != predicted)/length(predicted) # Overall error rate
     
344. round(100*table(actual, predicted, dnn=c("Actual", "Predicted"))/length(predicted))
     
345. 
     
346. 
     
347. ## ------------------------------------------------------------------------
     
348. plot(as.party(model))
     
349. 
     
350. 
     
351. ## ----out.lines=12--------------------------------------------------------
     
352. print(as.party(model))
     
353. 
     
354. 
     
355. ## ----c50-----------------------------------------------------------------
     
356. library(C50)
     
357. model <- C5.0(form, ds[train, vars])
     
358. 
     
359. 
     
360. ## ----c50_print, out.lines=NULL-------------------------------------------
     
361. model
     
362. 
     
363. 
     
364. ## ----out.lines=NULL------------------------------------------------------
     
365. C5imp(model)
     
366. 
     
367. 
     
368. ## ----c50_summary, out.lines=40-------------------------------------------
     
369. summary(model)
     
370. 
     
371. 
     
372. ## ----message=FALSE-------------------------------------------------------
     
373. predicted <- predict(model, ds[test, vars], type="prob")[,2]
     
374. riskchart(predicted, actual, risks)
     
375. 
     
376. 
     
377. ## ------------------------------------------------------------------------
     
378. predicted <- predict(model, ds[test, vars], type="class")
     
379. sum(actual != predicted)/length(predicted) # Overall error rate
     
380. round(100*table(actual, predicted, dnn=c("Actual", "Predicted"))/length(predicted))
     
381. 
     
382. 
     
383. ## ----c50_rules-----------------------------------------------------------
     
384. library(C50)
     
385. model <- C5.0(form, ds[train, vars], rules=TRUE)
     
386. 
     
387. 
     
388. ## ----c50_rules_print, out.lines=NULL-------------------------------------
     
389. model
     
390. 
     
391. 
     
392. ## ----out.lines=NULL------------------------------------------------------
     
393. C5imp(model)
     
394. 
     
395. 
     
396. ## ----c50_rules_summary, out.lines=40-------------------------------------
     
397. summary(model)
     
398. 
     
399. 
     
400. ## ----message=FALSE-------------------------------------------------------
     
401. predicted <- predict(model, ds[test, vars], type="prob")[,2]
     
402. riskchart(predicted, actual, risks)

复制代码

感谢分享

1. ## ------------------------------------------------------------------------
   
2. predicted <- predict(model, ds[test, vars], type="class")
   
3. sum(ds[test, target] != predicted)/length(predicted) # Overall error rate
   
4. round(100*table(ds[test, target], predicted, dnn=c("Actual", "Predicted"))/length(predicted))
   
5. 
   
6. 
   
7. ## ----build_regression_model, out.lines=NULL------------------------------
   
8. target <- "RISK_MM"
   
9. vars <- c(inputs, target)
   
10. form <- formula(paste(target, "~ ."))
    
11. (model <- rpart(formula=form, data=ds[train, vars]))
    
12. 
    
13. 
    
14. ## ----basic_plot_regression-----------------------------------------------
    
15. plot(model)
    
16. text(model)
    
17. 
    
18. 
    
19. ## ----basic_plot_uniform_regression---------------------------------------
    
20. plot(model, uniform=TRUE)
    
21. text(model)
    
22. 
    
23. 
    
24. ## ----basic_plot_extra_regressoin-----------------------------------------
    
25. plot(model, uniform=TRUE)
    
26. text(model, use.n=TRUE, all=TRUE, cex=.8)
    
27. 
    
28. 
    
29. ## ----fancy_plot_regression-----------------------------------------------
    
30. fancyRpartPlot(model)
    
31. 
    
32. 
    
33. ## ------------------------------------------------------------------------
    
34. prp(model)
    
35. 
    
36. 
    
37. ## ------------------------------------------------------------------------
    
38. prp(model, type=2, extra=101, nn=TRUE, fallen.leaves=TRUE,
    
39.     faclen=0, varlen=0, shadow.col="grey", branch.lty=3)
    
40. 
    
41. 
    
42. ## ----fig.width=14, out.width="\\textwidth"-------------------------------
    
43. class(model)
    
44. plot(as.party(model))
    
45. 
    
46. 
    
47. ## ------------------------------------------------------------------------
    
48. model <- ctree(formula=form, data=ds[train, vars])
    
49. 
    
50. 
    
51. ## ----out.lines=45--------------------------------------------------------
    
52. model
    
53. 
    
54. 
    
55. ## ------------------------------------------------------------------------
    
56. plot(model)
    
57. 
    
58. 
    
59. ## ----common_outtro, child="finale.Rnw", eval=TRUE------------------------
    
60. 
    
61. 
    
62. ## ----syinfo, child="sysinfo.Rnw", eval=TRUE------------------------------
    
63. 
    
64. ## ----echo=FALSE, message=FALSE-------------------------------------------
    
65. require(Hmisc)
    
66. pkg <- "knitr"
    
67. pkg.version <- installed.packages()[pkg, 'Version']
    
68. pkg.date <- installed.packages(fields="Date")[pkg, 'Date']
    
69. pkg.info <- paste(pkg, pkg.version, pkg.date)
    
70. 
    
71. rev <- system("bzr revno", intern=TRUE)
    
72. cpu <- system(paste("cat /proc/cpuinfo | grep 'model name' |",
    
73.                     "head -n 1 | cut -d':' -f2"), intern=TRUE)
    
74. ram <- system("cat /proc/meminfo | grep MemTotal: | awk '{print $2}'",
    
75.               intern=TRUE)
    
76. ram <- paste0(round(as.integer(ram)/1e6, 1), "GB")
    
77. user <- Sys.getenv("LOGNAME")
    
78. node <- Sys.info()[["nodename"]]
    
79. user.node <- paste0(user, "@", node)
    
80. gcc.version <- system("g++ -v 2>&1 | grep 'gcc version' | cut -d' ' -f1-3",
    
81.                       intern=TRUE)
    
82. os <- system("lsb_release -d | cut -d: -f2 | sed 's/^[ \t]*//'", intern=TRUE)

复制代码

什么东西那
书吗

kankan

看看！

谢谢分享

看看