-
Notifications
You must be signed in to change notification settings - Fork 0
/
Floor Model.R
310 lines (242 loc) · 8.61 KB
/
Floor Model.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
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
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
###Predicting Building
source("Preprocess.R")
## Start Parallel Processing ------------------
#detectCores() # Detect cores
#cl <- makeCluster(7) # Create Cluster
#registerDoParallel(cl) # Register Cluster
#getDoParWorkers() # Confirm number of core used
#stopCluster(cl) # Stop cluster
#remove(cl) # Remove cluster
###FUNCTIONS -----------------------------
# Returns wrong predictions
wrongPredRows <- function(dataset) {
errorList <- c()
x<-1
for (i in 1:nrow(dataset)){
if(dataset[i,]$pred != dataset[i,]$BUILDINGID){
errorList[x] <- i
x<-x+1
}
}
return(errorList)
}
removeRowsCols <- function(dataset) {
var1 <- dataset %>% select(-FLOOR, -BUILDINGID)
# List of WAPs with low/no signal
noSignalList <- apply(var1, 2,var) == 0
var1 <- var1[,!noSignalList]
# Remove rows (WAP) where all the values = 100 (WAP was not detected)
idx <- which(apply(var1[,1:(length(var1))], 1, function(x) all(x > 75 | x < 25)) == T)
if(!is_empty(idx)){
var1<-var1[-idx,]
}
#add buildingid, we need to remove the extra rows first
ExtractTrain <- select(dataset, FLOOR)
if(!is_empty(idx)){
ExtractTrain <- ExtractTrain[-idx,]
} else {
ExtractTrain <- as.integer(ExtractTrain$FLOOR)
}
var1$FLOOR <- ExtractTrain
return(var1)
}
## ALL BUILDINGS
#All Buildings to check our floor detection accuracy
BsubTraining <- subTraining
BsubTraining$FLOOR <- as.factor(BsubTraining$FLOOR)
levels(BsubTraining$FLOOR) <- c("F0", "F1", "F2", "F3", "F4")
#All Buildings subset for testing
BsubTesting <- subTesting
BsubTesting$FLOOR <- as.factor(BsubTesting$FLOOR)
levels(BsubTesting$FLOOR) <- c("F0", "F1", "F2", "F3", "F4")
## BUILDING 0
#Building 0 to check our floor detection accuracy
B0subTraining <- subTraining %>%
filter(BUILDINGID == "B0") #%>%
B0subTraining <- removeRowsCols(B0subTraining)
B0subTraining$FLOOR <- as.factor(B0subTraining$FLOOR)
levels(B0subTraining$FLOOR) <- c("F0", "F1", "F2", "F3")
#Building 0 subset for testing
B0subTesting <- subTesting %>%
filter(BUILDINGID == "B0")
B0subTesting$FLOOR <- as.factor(B0subTesting$FLOOR)
levels(B0subTesting$FLOOR) <- c("F0", "F1", "F2", "F3")
## BUILDING 1
#Building 1 to check our floor detection accuracy
B1subTraining <- subTraining %>%
filter(BUILDINGID == "B1") #%>%
B1subTraining <- removeRowsCols(B1subTraining)
B1subTraining$FLOOR <- as.factor(B1subTraining$FLOOR)
levels(B1subTraining$FLOOR) <- c("F0", "F1", "F2", "F3")
#Building 1 subset for testing
B1subTesting <- subTesting %>%
filter(BUILDINGID == "B1")
B1subTesting$FLOOR <- as.factor(B1subTesting$FLOOR)
levels(B1subTesting$FLOOR) <- c("F0", "F1", "F2", "F3")
## BUILDING 2
#Building 2 to check our floor detection accuracy
B2subTraining <- subTraining %>%
filter(BUILDINGID == "B2")
B2subTraining <- removeRowsCols(B2subTraining)
B2subTraining$FLOOR <- as.factor(B2subTraining$FLOOR)
levels(B2subTraining$FLOOR) <- c("F0", "F1", "F2", "F3", "F4")
#Building 2 subset for testing
B2subTesting <- subTesting %>%
filter(BUILDINGID == "B2")
B2subTesting$FLOOR <- as.factor(B2subTesting$FLOOR)
levels(B2subTesting$FLOOR) <- c("F0", "F1", "F2", "F3", "F4")
## Model Creation stage ------------------
#Splitting data into training set and testing set for cross validation
inTraining <- createDataPartition(BsubTraining$FLOOR, p = .80, list = FALSE)
training <- BsubTraining[inTraining,]
testing <- BsubTraining[-inTraining,]
###
## C5.0 Model
#10 fold cross validation
fitControl <- trainControl(method = "repeatedcv",
number = 3,
repeats = 1)
#Set seed to know the random order
set.seed(33)
c5grid <- expand.grid(.winnow = c(FALSE, TRUE),
.trials=c(30),
.model="trees" )
tic()
#train Linear Regression model
C5Fit <- train(FLOOR~., data = training,
method = "C5.0",
trControl=fitControl,
tuneLength = 5)
#tuneGrid = c5grid)
toc()
#c5 model
C5Fit
#Check validation set for accuracy on predicting the building
testing$pred <- predict(C5Fit, newdata = testing)
postResample(testing$pred, testing$FLOOR)
BsubTesting$pred <- predict(C5Fit, newdata = BsubTesting)
postResample(BsubTesting$pred, BsubTesting$FLOOR)
confusionMatrix(BsubTesting$pred, BsubTesting$FLOOR)
## Random Forest Model --------------
rfControl <- trainControl(method = "repeatedcv",
number = 3,
repeats = 1,
search="random")
rfGrid <- expand.grid(mtry=c(214,60))
#Set seed to know the random order
set.seed(33)
tic()
#Creating RF model
rfModel <- train(FLOOR~., data = training,
method = "rf",
trControl=rfControl,
tuneLength = 3,
#tuneGrid=rfGrid,
importance=T)
toc()
#view model
rfModel
rfModel<-readRDS("rfModel-Allfloors.rds")
#saveRDS(rfModel, file = "rfModel-Allfloors.rds")
testing$pred <- predict(rfModel, newdata = testing)
postResample(testing$pred, testing$FLOOR)
BsubTesting$pred <- predict(rfModel, newdata = BsubTesting)
postResample(BsubTesting$pred, BsubTesting$FLOOR)
confusionMatrix(BsubTesting$pred, BsubTesting$FLOOR)
## kNN Model
set.seed(1234)
knnControl = trainControl(method = "repeatedcv",
number = 3,
repeats = 1,
classProbs = TRUE,
summaryFunction = multiClassSummary)
tic()
knnModel <- train(FLOOR~. , data = training, method = "knn",
#preProcess = c("center","scale"),
trControl = knnControl,
#metric = "ROC",
tuneLength = 2)
toc()
# Summary of model
knnModel
#Check validation set for accuracy on predicting the building
testing$pred <- predict(knnModel, newdata = testing)
postResample(testing$pred, testing$FLOOR)
BsubTesting$pred <- predict(knnModel, newdata = BsubTesting)
postResample(BsubTesting$pred, BsubTesting$FLOOR)
confusionMatrix(BsubTesting$pred, BsubTesting$FLOOR)
## SVM Model
set.seed(63)
SVMControl <- trainControl(method = "repeatedcv",
number = 3,
repeats = 1)
#preProc = c("center", "scale", "range"))
SVMgrid <- expand.grid(C = c(1:20))
#Set seed to know the random order
tic()
SVMModel <- train(FLOOR ~., data = training, method = "svmLinear",
trControl=SVMControl,
#tuneGrid = SVMgrid
tuneLength = 25
)
toc()
#Run model and show output
SVMModel
#Check validation set for accuracy on predicting the building
testing$pred <- predict(SVMModel, newdata = testing)
postResample(testing$pred, testing$FLOOR)
BsubTesting$pred <- predict(SVMModel, newdata = BsubTesting)
postResample(BsubTesting$pred, BsubTesting$FLOOR)
#confusionMatrix(B1subTesting$pred, B1subTesting$FLOOR)
### GBM Model
fitControlGBM <- trainControl(method = "repeatedcv",
number = 3,
repeats = 1)
#Grid to define our own parameters for this classifier
gbmGrid <- expand.grid(interaction.depth = c(3,5),
n.trees = (5)*50,
shrinkage = c(0.1),
n.minobsinnode = 20)
#Set seed to know the random order
set.seed(998)
#Train GBM Regression model
tic()
gbmFit <- train(FLOOR~., data = training,
method = "gbm",
trControl = fitControlGBM,
verbose = FALSE,
tuneLength = 5)
#tuneGrid = gbmGrid)
toc()
#training results
gbmFit
#Check validation set for accuracy on predicting the building
testing$pred <- predict(gbmFit, newdata = testing)
postResample(testing$pred, testing$FLOOR)
BsubTesting$pred <- predict(gbmFit, newdata = BsubTesting)
postResample(BsubTesting$pred, BsubTesting$FLOOR)
## Error check --------------
errorList <- wrongPredRows(subTesting)
#get rows with mistakes
temp <- subTesting %>%
slice(errorList)
#leave only WAPs
temp2 <- select(temp, starts_with("WAP"))
# List of WAPs with low/no signal
noSignalList <- apply(temp2, 2,var) == 0
temp2 <- temp2[,!noSignalList]
temp2<- cbind(temp2, temp[283:287])
# Building Preview 3D
BsubTesting %>%
plot_ly(x = ~LONGITUDE,
y = ~LATITUDE,
z = ~as.factor(FLOOR),
color = ~pred,
colors = c("green", "orange", "yellow", "pink", "purple")) %>%
add_markers() %>%
layout(title = "Building Preview",
scene = list(xaxis = list(title = "Longitude"),
yaxis = list(title = "Latitude"),
zaxis = list(title = "Floor")))
errorcheck <- subTesting %>%
filter(BUILDINGID == "B0" & FLOOR == 3 & LATITUDE > 4864951 & LATITUDE < 4864999)