-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy path13_ffs_haralick_turnover_all_predictors.R
184 lines (146 loc) · 6.76 KB
/
13_ffs_haralick_turnover_all_predictors.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
library(gpm)
library(CAST)
setwd("/media/sd19006/data/users/iotte/R-Server/hyperspectral/clean")
## Read files and build GPM object
filepath_gpm = "/media/sd19006/data/users/iotte/R-Server/hyperspectral/clean/"
mrg_tbl = read.table(paste0(filepath_gpm,
"veg_beta_diversity_all_taxa_ffs_TO_NE_AC_df.csv"),
header = TRUE, sep = ";", dec = ",")
mrg_tbl$SelCat = substr(mrg_tbl$plotID, 1, 3)
mrg_tbl$SelNbr = as.numeric(substr(mrg_tbl$plotID, 4, 4))
col_selector = which(names(mrg_tbl) %in% c("SelCat", "SelNbr"))
col_diversity = seq(grep("ants_jtu_NMDS1", names(mrg_tbl)),
grep("rosids_jac_NMDS2", names(mrg_tbl)))
col_predictors = seq(grep("mean_Carter", names(mrg_tbl)),
grep("sd_TGI", names(mrg_tbl)))
col_meta = seq(length(mrg_tbl))[-c(col_selector, col_diversity, col_predictors)]
#######################
setwd("/media/sd19006/data/users/iotte/R-Server/hyperspectral/clean")
#ele = read.csv2("/media/sd19006/data/users/iotte/R-Server/hyperspectral/clean/Biodiversity_Data_Marcel.csv")
#ele = ele[-c(6:11),]
#ele = ele[-c(7,24),]
## Read files and build GPM object
filepath_gpm = "/media/sd19006/data/users/iotte/R-Server/hyperspectral/clean/"
mrg_tbl = read.table(paste0(filepath_gpm, "rs_veg_hara_species_turnover_TO_NE_AC_df.csv"),
header = TRUE, sep = ";", dec = ",")
test100 <- colSums(mrg_tbl[,c(2:1687)])
test100_4 <- which(is.na(test100))
test101 <- as.numeric(test100_4)
mrg_tbl_no = seq(2:1687)
mrg_tbl_tst <- mrg_tbl[,][-which(mrg_tbl_no %in% test101)]
any(is.na(mrg_tbl_tst))
which(is.na(mrg_tbl_tst))
test100 <- colSums(mrg_tbl_tst[,c(2:1143)])
test100_4 <- which(is.na(test100))
test101 <- as.numeric(test100_4)
mrg_tbl_res <- mrg_tbl_tst[,-c(759, 761, 783, 785, 807, 809, 831, 833, 855, 857, 879, 881,
903, 905, 927, 929, 951, 953, 975, 977, 999, 1001, 1023,
1025, 1047, 1049, 1071, 1073, 1095, 1097, 1119, 1121)]
any(is.na(mrg_tbl_res[,2:1111]))
mrg_tbl = mrg_tbl_res
mrg_tbl$SelCat = substr(mrg_tbl$plotID, 1, 3)
mrg_tbl$SelNbr = as.numeric(substr(mrg_tbl$plotID, 4, 4))
col_selector = which(names(mrg_tbl) %in% c("SelCat", "SelNbr"))
col_diversity = seq(grep("ants_jtu_NMDS1", names(mrg_tbl)),
grep("rosids_jac_NMDS2", names(mrg_tbl)))
#col_diversity = c(col_diversity[1]-1, col_diversity)
col_predictors = seq(grep("rs_sd_miv", names(mrg_tbl)),
grep("SR_smpl_inverse_difference_moment_30_sd", names(mrg_tbl)))
#col_precitors = col_precitors[-which(col_precitors %in% 28)] # "IQR_M_OSAVI" || "min_GDVI_4"
col_meta <- seq(length(mrg_tbl))[-c(col_selector, col_diversity, col_precitors)]
#######################
meta = createGPMMeta(mrg_tbl, type = "input",
selector = col_selector,
response = col_diversity,
predictor = col_predictors,
meta = NULL)
mrg_tbl_gpm = gpm(mrg_tbl, meta, scale = TRUE)
#mrg_tbl_gpm = createIndexFolds(x = mrg_tbl_gpm, nbr = 1)
## Iterate over all response variables
responses = mrg_tbl_gpm@meta$input$RESPONSE_FINAL
mrg_tbl_gpm_list = lapply(responses, function(rsp){
mrg_tbl_gpm@meta$input$RESPONSE_FINAL = rsp
## Remove NAs and compute resamples
mrg_tbl_gpm@data$input = mrg_tbl_gpm@data$input[complete.cases(mrg_tbl_gpm@data$input[, mrg_tbl_gpm@meta$input$RESPONSE_FINAL]), ]
#mrg_tbl_gpm = splitMultRespLSO(x = mrg_tbl_gpm, nbr = 1)
mrg_tbl_gpm = createIndexFolds(x = mrg_tbl_gpm, nbr = 1) #nested_cv = FALSE)
mrg_tbl_gpm_ffs_model = trainModel(x = mrg_tbl_gpm,
metric = "RMSE",
n_var = NULL,
mthd = "pls",
mode = "ffs",
seed_nbr = 11,
cv_nbr = 5,
var_selection = "indv")
filepath = paste0(filepath_gpm, rsp, ".rds")
saveRDS(mrg_tbl_gpm_ffs_model, file = filepath)
})
saveRDS(mrg_tbl_gpm, file = "mrg_tbl_gpm_ffs_all_taxa.rds")
filepath = paste(filepath_gpm, "gpm_ffs_40", sep = "")
## Combine models in gpm objct
models = list.files(filepath, pattern = glob2rx("*.rds"), full.names = TRUE)
mrg_tbl_gpm_model = readRDS(models[1])
for(i in seq(2, length(models))){
temp = readRDS(models[i])
mrg_tbl_gpm_model@model$pls_rfe[[i]] = temp@model$pls_ffs[[1]]
}
saveRDS(mrg_tbl_gpm_model, file = paste0(filepath, "gpm_ffs_40_2018-05-11.rds"))
#filepath = paste(filepath_gpm, "residuen", sep = "")
#mrg_tbl_res_gpm_model <- readRDS(paste0(filepath_gpm, "gpm_pci_hara_res_model_pls_2018-03-12.rds"))
mrg_tbl_gpm_model <- readRDS(paste0(filepath_gpm, "gpm_ffs_40_2018-05-11.rds"))
compVarImp <- function(models, scale = FALSE){
lapply(models, function(x){
vi_species1 <- lapply(x, function(y){
# vi <- varImp(y$model$fit, scale = FALSE) #war: var_Imp(y$model$fit, scale = FALSE)
if(inherits(y$model, "try-error")){
NULL
} else {
vi <- caret::varImp(y$model)
class(vi)
if(inherits(vi, "varImp.train")){
vi = vi$importance
}
if(scale == TRUE){
vi <- vi / max(vi)
}
variables <- rownames(vi)
vi <- data.frame(RESPONSE = y$response,
VARIABLE = variables,
IMPORTANCE = vi$Overall)
}
})
vi_species <- do.call("rbind", vi_species1)
if(is.null(vi_species)){
vi <- NULL
} else {
#return(vi_species)
#}
#})
#}
#vi_count <- vi_species %>% count(VARIABLE)
vi_count <- vi_species %>% dplyr::count(VARIABLE)
#vi_mean <- vi_species %>% group_by(RESPONSE) %>% summarise(avg = mean(IMPORTANCE))
vi_mean <- ddply(vi_species, "VARIABLE", summarise, mean = mean(IMPORTANCE))
vi <- merge(vi_count, vi_mean)
vi$RESPONSE <- vi_species$RESPONSE[1]
vi <- vi[order(vi$mean, decreasing = TRUE), ,drop = FALSE]
}
return(vi)
})
}
#mrg_tbl_gpm_model = mrg_tbl_res_gpm_model
var_imp <- compVarImp(mrg_tbl_gpm_model@model[[2]], scale = FALSE)
var_imp_scale <- compVarImp(mrg_tbl_gpm_model@model[[2]], scale = TRUE)
tst = do.call(rbind, var_imp_scale)
library(stringr)
tst2 = str_sort(tst$VARIABLE)
unique(tst2)
var_imp_plot <- plotVarImp(var_imp)
var_imp_heat <- plotVarImpHeatmap(var_imp_scale, xlab = "Species", ylab = "Band")
tstat <- compRegrTests(mrg_tbl_gpm_model@model[[2]])
#tstat_mean <- merge(tstat[[1]], mrg_tbl_res_gpm_model@model[[1]],
# by.x = "Response", by.y="names")
overview = aggregate(tstat$r_squared, by = list(tstat$model_response), mean)
colnames(overview) = c("Species Richness", "r.sq")
#overview$r.sq_residuen_rd = round(overview$r.sq_residuen, 3)
overview[order(overview$r.sq),]