trendsegment/methods.R at master · hmaeng/trendsegment · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199

#######################################
#   Methods
#######################################

ts <- function(x, thr, p, bal, method, beta, ...){

  tic <- proc.time()
  object <- trendsegment(x, th.const = thr, p = p, bal = bal)
  toc <- proc.time()

  list(fit = object$est, cpts=object$cpt, elapsed=(toc-tic)[3])

}

not.sic <- function(x){

  tic <- proc.time()
  object <- not(x, method = "not", contrast = "pcwsLinMean", parallel = FALSE) # give an error in noiseless input
  cpts <- features(object, penalty="sic")$cpt
  fit <- predict(object, cpt=cpts)
  toc <- proc.time()

  list(fit = fit, cpts=cpts, elapsed=(toc-tic)[3])
}

id <- function(x){ # adjusted cpts after fitting
  # e.g. x <- c(0:9, 5, 10:0)+rnorm(22)/3  gives three consecutive change-points
  tic <- proc.time()
  #object <- ID(x, contrast ="slope", ht=F) # gives a weird cpts in noiseless input
  object <- ID(x, contrast ="slope", ht=F, lambda=1) # this does not make any difference compared to the upper one in MODEL 5-7 which contain spikes
  if(length(object$cpt)==1 && object$cpt==c(0)){
    cpts <- c()
  } else{
    cpts <- object$cpt
  }

  ### adjust change-points as it gives one more change-point for each estimated change-points by finding the point which gives nonzero second derivative.
  if(length(cpts)>2){
    # 1) searching spikes only
    threeeach <- cbind(cpts[1:(length(cpts)-2)], cpts[2:(length(cpts)-1)], cpts[3:(length(cpts))])
    dthr <- t(apply(threeeach,1,diff))
    spike <- which(dthr[,1]==1 & dthr[,2]==1)
    # 2) searching jump only
    twoeach <- cbind(cpts[1:(length(cpts)-1)], cpts[2:(length(cpts))])
    jump <- which(t(apply(twoeach,1,diff))==1)
    jump <-jump[!jump %in% c(spike, spike+1)] # get rid of the cpts for spikes
    # removing redundunt
    if(length(spike)>0 | length(jump)>0){
      cpts <- cpts[-unique(c(spike+2, jump+1))]
    } else{
      cpts <- cpts
    }
  } else if(length(cpts)==2 && diff(cpts)[1]==1){
    cpts <- cpts[-2]
  } else {
    cpts <- cpts
  }

  toc <- proc.time()

  list(fit = object$fit, cpts=cpts, elapsed=(toc-tic)[3])

}

tf <- function(x){ # adjusted cpts after fitting
  # e.g. x <- c(rep(0,9), 2, 10:0) gives three consecutive cpts
  tic <- proc.time()
  object <-  trendfilter(y=x, ord=1)

  tf.cv <- cv.trendfilter(object)
  cpts <- which(abs(diff(object$fit[,tf.cv$i.min], differences=2)) > sqrt(.Machine$double.eps))+1

  ### adjust change-points as it gives one more change-point for each estimated change-points by finding the point which gives nonzero second derivative.
  if(length(cpts)>2){
    # 1) searching spikes only
    threeeach <- cbind(cpts[1:(length(cpts)-2)], cpts[2:(length(cpts)-1)], cpts[3:(length(cpts))])
    dthr <- t(apply(threeeach,1,diff))
    spike <- which(dthr[,1]==1 & dthr[,2]==1)
    # 2) searching jump only
    twoeach <- cbind(cpts[1:(length(cpts)-1)], cpts[2:(length(cpts))])
    jump <- which(t(apply(twoeach,1,diff))==1)
    jump <-jump[!jump %in% c(spike, spike+1)] # get rid of the cpts for spikes
    # removing redundunt
    if(length(spike)>0 | length(jump)>0){
      cpts <- cpts[-unique(c(spike+2, jump+1))]
    } else{
      cpts <- cpts
    }
  } else if(length(cpts)==2 && diff(cpts)[1]==1){
    cpts <- cpts[-2]
  } else {
    cpts <- cpts
  }

  toc <- proc.time()

  list(fit = object$fit[,tf.cv$i.min], cpts=cpts, elapsed=(toc-tic)[3])

}

cpop <- function(x){ # adjusted cpts after fitting
  # e.g. x <- c(0:9, 5, 10:0)+rnorm(22)/3 returns three consecutive cpts
  tic <- proc.time()
  sig.hat <- mad(diff(diff(x)))/sqrt(6) # give an error in noiseless input
  object <- CPOP.run(x/sig.hat)
  cpts <- object$cpt
  fit <- object$f*sig.hat

  ### adjust change-points as it gives one more change-point for each estimated change-points by finding the point which gives nonzero second derivative.
  if(length(cpts)>2){
    # 1) searching spikes only
    threeeach <- cbind(cpts[1:(length(cpts)-2)], cpts[2:(length(cpts)-1)], cpts[3:(length(cpts))])
    dthr <- t(apply(threeeach,1,diff))
    spike <- which(dthr[,1]==1 & dthr[,2]==1)
    # 2) searching jump only
    twoeach <- cbind(cpts[1:(length(cpts)-1)], cpts[2:(length(cpts))])
    jump <- which(t(apply(twoeach,1,diff))==1)
    jump <-jump[!jump %in% c(spike, spike+1)] # get rid of the cpts for spikes
    # removing redundunt
    if(length(spike)>0 | length(jump)>0){
      cpts <- cpts[-unique(c(spike+2, jump+1))]
    } else{
      cpts <- cpts
    }
  } else if(length(cpts)==2 && diff(cpts)[1]==1){
    cpts <- cpts[-2]
  } else {
    cpts <- cpts
  }

  toc <- proc.time()

  list(fit = fit, cpts=cpts, elapsed=(toc-tic)[3])
}

bup <- function(x, max.err=300){

  tic <- proc.time()

  left_x <- seq(1, length(x)-1, by=2)
  right_x <- left_x + 1
  right_x[length(right_x)] <- length(x)
  number_of_segments <- length(left_x)

  segment <- cbind(left_x, right_x, Inf)

  for(i in 1:(number_of_segments-1)){
    lmfit <- lm(x[c(segment[i, 1]:segment[i+1, 2])]~ c(segment[i, 1]:segment[i+1, 2]))
    segment[i, 3] <- sum((lmfit$residuals)^2) # sse
  }

  while(min(segment[,3]) < max.err){ # max.err

    i <- which.min(segment[,3])

    if(i==1){
      lmfit <- lm(x[c(segment[i, 1]:segment[i+2, 2])]~ c(segment[i, 1]:segment[i+2, 2]))
      segment[i, 3] <- sum((lmfit$residuals)^2) # sse

      segment[i, 2] <- segment[i+1, 2]
      segment <- segment[-(i+1),,drop=F]
    } else if(i>1 && i < dim(segment)[1]-1){
      lmfit <- lm(x[c(segment[i, 1]:segment[i+2, 2])]~ c(segment[i, 1]:segment[i+2, 2]))
      segment[i, 3] <- sum((lmfit$residuals)^2) # sse

      segment[i, 2] <- segment[i+1, 2]
      segment <- segment[-(i+1),,drop=F]

      i <- i-1
      lmfit <- lm(x[c(segment[i, 1]:segment[i+1, 2])]~ c(segment[i, 1]:segment[i+1, 2]))
      segment[i, 3] <- sum((lmfit$residuals)^2) # sse
    } else{
      segment[i, 2] <- segment[i+1, 2]
      segment[i, 3] <- Inf
      segment <- segment[-(i+1),,drop=F]

      i <- i-1
      lmfit <- lm(x[c(segment[i, 1]:segment[i+1, 2])]~ c(segment[i, 1]:segment[i+1, 2]))
      segment[i, 3] <- sum((lmfit$residuals)^2) # sse
    }

  }

  ### change-points
  cpt <- c(segment[-dim(segment)[1], 2])
  ### estimated curve
  est <- rep(NA, length(x))
  for(i in 1:dim(segment)[1]){
    lmfit <- lm(x[segment[i,1]:segment[i,2]]~c(segment[i,1]:segment[i,2]))
    est[c(segment[i,1]:segment[i,2])] <- lmfit$fitted.values
  }


  toc <- proc.time()

  list(fit=est, cpts=cpt, elapsed=(toc-tic)[3])

}