forked from lkaihua/Bipeline
-
Notifications
You must be signed in to change notification settings - Fork 0
/
segMerge.R
171 lines (142 loc) · 4.44 KB
/
segMerge.R
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
# The function plotKMeans returns a plot of clustered data using k-means
#
# Author: Kaihua Liu
###########################################################################################################
source('LSDD.R')
source('LSDDsegmentation.R')
segMerge <- function(data, segResults, segLSDDPars, throttle = 100){
# result
newSeg <- data.frame()
# book keeping
mergedRight <- FALSE
for(i in 1:nrow(segResults)){
# check if previous seg was merged to the right
if(mergedRight == TRUE) {
mergedRight <- FALSE
next
}
# sorry you are too small we have to kill you
if(segResults[i, "segEnd"] - segResults[i, "segStart"] + 1 <= throttle){
# if it is the first seg or the last seg
# it only can be merged to one direction
# so stop calling the function
if(i == 1){
mergeDirection <- 'right'
}
else if(i == nrow(segResults)){
mergeDirection <- 'left'
}
# use LSDD fast function to calculate similarity
else {
mergeDirection <- LSDDCompare(
L = segResults[i-1, ],
G = segResults[i, ],
R = segResults[i+1, ],
data = data,
segLSDDPars = segLSDDPars
)
}
if(mergeDirection == "left"){
# merge to left
# edit last seg
newSeg[nrow(newSeg), "segEnd"] <- segResults[i, "segEnd"]
}else{
# merge to right
# push to new vector
newSeg <- rbind(newSeg,
list(
"segStart" = segResults[i,"segStart"],
"segEnd" = segResults[i+1, "segEnd"]
)
)
# skip seg to be deal in next loop
mergedRight <- TRUE
}
}
# You are long enough to pass
else{
newSeg <- rbind(newSeg, segResults[i,])
}
}
newSeg <- unique(newSeg)
return(newSeg)
}
#######################################
######### LSDDCompare ##########
#######################################
#
# |------------|-----|-----------|
# L G R
# Compare Gap with two different time sequences
# And return the one with higher similarity
#
#######################################
# Input:
#
# L, G, R:three time sequences to deal with. list("segStart" = , "segEnd" = )
# G is the gap (seg size below the minimum)
#
# data: dataset
#
# segLSDDPars: sigma and lambdas matrices
#######################################
# Output:
#
# "left"/"right", the side to merge
LSDDCompare <- function(L, G, R, data, segLSDDPars){
sum_LSDD_G2L <- 0
sum_LSDD_G2R <- 0
# under some extrme conditions
# the size of G could be 1, which means only 1 data point
# in this segments,
# but LSDD requires at least 2 data points to carry on.
# so just merge G to the shorter side
if(
G$segEnd - G$segStart > 1
&& L$segEnd - L$segStart > 1
&& R$segEnd - R$segStart > 1
){
for(i in 1:ncol(data)){
# 1 x n matrix
Gmatrix <- t(matrix(data[G$segStart:G$segEnd, i]))
Lmatrix <- t(matrix(data[L$segStart:L$segEnd, i]))
Rmatrix <- t(matrix(data[R$segStart:R$segEnd, i]))
# cat(dim(Gmatrix), dim(Lmatrix), dim(Rmatrix))
sum_LSDD_G2L <- sum(sum_LSDD_G2L, LSDDfast(
X1 = Gmatrix,
X2 = Lmatrix,
sigma = segLSDDPars[i, "sigma"], #sigma
lambda = segLSDDPars[i, "lambda"] #lambda
))
sum_LSDD_G2R <- sum(sum_LSDD_G2R, LSDDfast(
X1 = Gmatrix,
X2 = Rmatrix,
sigma = segLSDDPars[i, "sigma"], #sigma
lambda = segLSDDPars[i, "lambda"] #lambda
))
}
}
# cat("Similary to left is: ", sum_LSDD_G2L, "\n")
# cat("Similary to right is: ", sum_LSDD_G2R, "\n")
if(sum_LSDD_G2R < sum_LSDD_G2L){
result <- 'right'
}else if(sum_LSDD_G2L < sum_LSDD_G2R){
result <- 'left'
}else{
# this is equal, very rare situation, then merge with shorter segResults
if(L$segEnd - L$segStart < R$segEnd - R$segStart){
result <- 'left'
}
else{
result <- 'right'
}
}
# cat("Merge to ", result, "\n")
result
}
####### test #######
# data <- data.frame( read.csv('./data/test5000.csv'))
# data <- data[-1]
# segResults <- data.frame( read.csv('./data/segs5000.csv'))
# segLSDDPars <- data.frame( read.csv('./data/pars5000.csv'))
# segMerge(data = data, segResults = segResults, segLSDDPars = segLSDDPars, throttle = 100)