-
Notifications
You must be signed in to change notification settings - Fork 7
/
Copy pathcalculate_accuracy.R
366 lines (303 loc) · 17.4 KB
/
calculate_accuracy.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
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
library(ggplot2)
library(gridExtra)
library(dplyr)
library(data.table)
library(tidyr)
source("accuracy_functions.R")
# get same data as used for allocations
source_lines("calculate_allocations.R", 7:26) # careful!
#rm(survey_points_raw)
iters <- list.files("A:/1_UNSW/0_data/Dharawal_project/iters/", pattern = ".rds", full.names = T)
big_list <- lapply(iters, FUN = function(x){readRDS(x)})
big_list_image <- readRDS("big_list_image.rds")
#readRDS("A:/1_UNSW/0_data/Dharawal_project/big_list_alldat.rds")
# decide what results to extract - be VERY careful, and examine the data frame well
get_this <- rbind(
# for lda classifier
data.frame(
scenario = c(rep("boot", 3),
rep("rrcv", 24),
rep("kfold", 12)),
type = c(rep("boot", 3),
rep(names(big_list[[1]][["rrcv"]]), each = 3),
rep(names(big_list[[1]][["kfold"]]), each = 3)),
method = c(rep(c("train_lda", "test_lda", "true_lda"), 13)),
tt = c(rep(c("train", "test", "true"), 13)),
stringsAsFactors = F),
# # for knn classifier
# data.frame(
# scenario = c(rep("boot", 2),
# rep("rrcv", 16),
# rep("kfold", 8)),
# type = c(rep("boot",2),
# rep(names(big_list[[1]][["rrcv"]]), each = 2),
# rep(names(big_list[[1]][["kfold"]]), each = 2)),
# method = c(rep(c("test_knn", "true_knn"), 13)),
# tt = c(rep(c("test", "true"), 13)),
# stringsAsFactors = F),
# for rf classifier
data.frame(
scenario = c(rep("boot", 3),
rep("rrcv", 24),
rep("kfold", 12)),
type = c(rep("boot", 3),
rep(names(big_list[[1]][["rrcv"]]), each = 3),
rep(names(big_list[[1]][["kfold"]]), each = 3)),
method = c(rep(c("train_rf", "test_rf", "true_rf"), 13)),
tt = c(rep(c("train", "test", "true"), 13)),
stringsAsFactors = F)
)
get_this_image <- data.frame(
scenario = c(rep("boot", 2),
rep("rrcv", 16),
rep("kfold", 8)),
type = c(rep("boot", 2),
rep(names(big_list_image[[1]][["rrcv"]]), each = 2),
rep(names(big_list_image[[1]][["kfold"]]), each = 2)),
method = c(rep(c("image_lda", "image_rf"), 13)),
tt = c(rep(c("test_lda", "test_rf"), 13)),
stringsAsFactors = F)
# calculate stats ---------------------------------------------------------
# for image_preds
image_results <- collect_image_iteration(1, get_this_image, big_list_image, survey_points) %>%
na.omit()
saveRDS(image_results, file="image_results.rds")
metric_results <- rbindlist(lapply(
X = 1:length(big_list),
FUN = collect_one_iteration,
get_this, big_list, survey_points
))
saveRDS(metric_results, file="metric_results.rds")
# metric_results_alldat <- rbindlist(lapply(
# X = 1:length(big_list_alldat),
# FUN = collect_one_iteration,
# get_this_all, big_list_alldat, survey_points
# ))
# metric_results_alldat <- filter(metric_results_alldat, scenario == "alldat")
# saveRDS(metric_results_alldat, file="metric_results_alldat.rds")
#readRDS("metric_results_full.rds")
#readRDS("metric_results_alldat.rds")
#metric_results <- rbind(metric_results, metric_results_alldat)
#saveRDS(metric_results, file="metric_results.rds")
# plots -------------------------------------------------------------------------
metric_results <- readRDS("metric_results.rds")
image_results <- readRDS("image_results.rds")
metric_results_long <- prettify_results(metric_results)
image_results_long <- prettify_results_image(image_results)
##### main paper figures
# fig. 2 (accuracy resutls)
fig2 <- metric_results_long %>%
filter(sample_origin %in% c("test"),
sample_structure %in% c("bootstrap", "random","block", "class", "class-space"),
model == "max-likelihood",
metric %in% c("perc_agr", "kappa", "entropy", "purity", "quant_dis", "alloc_dis")) %>%
mutate(metric = recode(metric, "perc_agr" = "overall accuracy", "quant_dis" = "quant. dis.", "alloc_dis" = "alloc. dis."),
sample_structure = recode(sample_structure, "class-space" = "class & space")) %>%
ggplot(., aes(y = value)) +
geom_violin(aes(x = sample_origin, fill = sample_fraction), scale = "area", draw_quantiles = c(0.05,0.5,0.9), lwd=0.25) +
scale_fill_manual("Resampling design", values = c("#969696", "#d55e00", "#f0e442", "#56b4e9")) +
ylab("Accuracy metric value") + xlab("Sampling stratification design") +
theme_bw() + theme(axis.text.x = element_blank(), strip.text = element_text(size = 14),
axis.ticks.x = element_blank(), axis.title = element_text(size = 14),
strip.background = element_rect(colour = NA, fill = NA)) +
facet_grid(metric ~ sample_structure, scales = "free", space = "free", drop = T)
# manual mucking with displays
fig2G <- ggplotGrob(fig2)
fig2G$widths[c(4)] <- unit(1,"null")
fig2G$widths[c(6,8,10,12)] <- unit(3,"null")
fig2G$heights[c(7)] <- unit(1,"null")
fig2G$heights[c(9,11,13,15,17)] <- unit(0.4,"null")
# grid.newpage()
# grid.draw(fig2G)
ggsave(plot = fig2G, filename = paste0("plots/figure2.pdf"), device = "pdf", width = 18, height = 12.5)
# fig. 3 (area estimates)
fig3dat <- image_results_long %>%
filter(model == "max-likelihood") %>%
select(-class, -user_prod, -sample_origin, -model) %>%
mutate(metric = recode_factor(metric, "Banksia" = "Banksia Thicket", "Eucalypt" = "Eucalypt Woodland",
"Teatree" = "Tea Tree Thicket", "Wetheath" = "Wet Heath"),
sample_structure = recode(sample_structure, "class-space" = "class & space"),
value = value * 100) # % for prettier numbers on plot
fig3 <- ggplot(fig3dat, aes(y = value)) +
geom_violin(aes(x = sample_structure, fill = sample_fraction), scale = "area", draw_quantiles = c(0.05,0.5,0.9), lwd=0.25) +
scale_fill_manual("Resampling design", values = c("#969696", "#d55e00", "#f0e442", "#56b4e9")) +
scale_y_continuous(breaks = pretty_breaks) +
ylab("Proportion of mapped area (%)") +
xlab("Sampling stratification design") +
theme_bw() + theme(axis.text.x = element_blank(), strip.text = element_text(size = 14),
axis.ticks.x = element_blank(), axis.title = element_text(size = 14),
strip.background = element_rect(colour = NA, fill = NA)) +
facet_grid(metric ~ sample_structure, scales = "free", space = "free", drop = T)
# manual mucking with displays
fig3G <- ggplotGrob(fig3)
fig3G$widths[c(4)] <- unit(1,"null")
fig3G$widths[c(6,8,10,12)] <- unit(3,"null")
fig3G$heights[c(7,9,11,13)] <- unit(1,"null")
# grid.newpage()
# grid.draw(fig2G)
ggsave(plot = fig3G, filename = paste0("plots/figure3.pdf"), device = "pdf", width = 18, height = 12.5)
##### supplementary material plots
# fig. X (user producer?)
#stats first
# looks at mean/CI/min-max summaries
metric_stats <- metric_results_long %>%
group_by(model, sample_structure, sample_fraction, sample_origin, metric) %>%
summarise(mean = mean(value),
median = median(value),
upper = quantile(value, 0.95),
lower = quantile(value, 0.05),
max = max(value),
min = min(value)) %>%
mutate_at(vars(mean:min), funs(round(.,3))) %>%
ungroup() %>%
mutate(sample_structure = recode(sample_structure, "bootstrap" = "BS", "class-space" = "class & space"))
# User/prod table supp plots
user_prod_names <- c("bt_prod", "ew_prod", "ttt_prod", "wh_prod", "bt_user", "ew_user", "ttt_user", "wh_user")
user_prod_table <- metric_stats %>%
filter(sample_origin == "test",
metric %in% user_prod_names)
user_prod_table$class <- NA
user_prod_table$class[grep("bt", user_prod_table$metric)] <- "Banksia"
user_prod_table$class[grep("ew", user_prod_table$metric)] <- "Eucalypt"
user_prod_table$class[grep("ttt", user_prod_table$metric)] <- "Tea tree"
user_prod_table$class[grep("wh", user_prod_table$metric)] <- "Wet heath"
user_prod_table$user_prod <- NA
user_prod_table$user_prod[grep("user", user_prod_table$metric)] <- "user"
user_prod_table$user_prod[grep("prod", user_prod_table$metric)] <- "producer"
#write.csv(user_prod_table, file = "plots/user_prod_table.csv")
user_prod_ml <- user_prod_table %>%
filter(sample_origin %in% c("test"),
sample_structure %in% c("BS", "random","block", "class", "class & space"),
model == "max-likelihood",
metric %in% c("bt_user", "ew_user", "ttt_user", "wh_user","perc_agr", "bt_prod", "ew_prod", "ttt_prod", "wh_prod")) %>%
ggplot(., aes(y = median)) +
geom_errorbar(aes(x = sample_fraction, ymin = lower, ymax = upper), width = 0) +
geom_point(aes(x = sample_fraction)) +
#geom_boxplot(aes(x = sample_fraction)) +
scale_y_continuous(breaks = c(0,0.5,1)) +
#scale_fill_manual("Resampling design", values = c("#969696", "#d55e00", "#f0e442", "#56b4e9")) +
#scale_colour_manual("Resampling design", values = c("#969696", "#d55e00", "#f0e442", "#56b4e9")) +
ylab("Accuracy value") + xlab("Sampling stratification design") +
theme_bw() + theme(strip.text = element_text(size = 12), axis.title = element_text(size=14)) +
facet_grid(class + user_prod ~ sample_structure, scales = "free_x", space = "free", drop = T)
ggsave(plot = user_prod_ml, filename = paste0("plots/user_prod_ml.png"), device = "png", width = 11, height = 8)
user_prod_rf <- user_prod_table %>%
filter(sample_origin %in% c("test"),
sample_structure %in% c("BS", "random","block", "class", "class & space"),
model == "random-forest",
metric %in% c("bt_user", "ew_user", "ttt_user", "wh_user","perc_agr", "bt_prod", "ew_prod", "ttt_prod", "wh_prod")) %>%
ggplot(., aes(y = median)) +
geom_errorbar(aes(x = sample_fraction, ymin = lower, ymax = upper), width = 0) +
geom_point(aes(x = sample_fraction)) +
scale_y_continuous(breaks = c(0,0.5,1)) +
#scale_fill_manual("Resampling design", values = c("#969696", "#d55e00", "#f0e442", "#56b4e9")) +
#scale_colour_manual("Resampling design", values = c("#969696", "#d55e00", "#f0e442", "#56b4e9")) +
ylab("Accuracy value") + xlab("Sampling stratification design") +
theme_bw() + theme(strip.text = element_text(size = 12), axis.title = element_text(size=14)) +
facet_grid(class + user_prod ~ sample_structure, scales = "free_x", space = "free", drop = T)
ggsave(plot = user_prod_rf, filename = paste0("plots/user_prod_rf.png"), device = "png", width = 11, height = 8)
# supp plots for full ML/RF results
plot_by_structure(metric_results_long, "max-likelihood", origins = c("test", "true"))
plot_by_structure(metric_results_long, "random-forest", origins = c("test", "true"))
#plot_by_structure(metric_results_long, "nearest-n", origins = c("train", "test", "true"))
#plot_by_structure(metric_results_long, "max-likelihood", origins = c("test"), suffix = "-test")
#plot_by_model(metric_results_long, c("max-likelihood","random-forest", "nearest-n"), origins = c("test"), suffix = "ml-nn-rf-test")
plot_user_prod(metric_results_long, c("max-likelihood"), suffix = "ml-user-prod-test")
plot_user_prod(metric_results_long, c("max-likelihood"), origins = c("train"), suffix = "ml-user-prod-train")
# plot_user_prod(metric_results_long, c("max-likelihood"), origins = c("true"), suffix = "ml-true-user-prod")
plot_by_structure(metric_results_long, "max-likelihood", origins = c("true","test"),
metrics = c("bt_user", "ew_user", "ttt_user", "wh_user"),
suffix = "-user", scales = "free_x")
plot_by_structure(metric_results_long, "max-likelihood", origins = c("true","test"),
metrics = c("bt_prod", "ew_prod", "ttt_prod", "wh_prod"),
suffix = "-producer", scales = "free_x")
# plot_by_structure(metric_results_long, "max-likelihood", origins = c("train","test","true"))
# plot_by_structure(metric_results_long, "random-forest", origins = c("all", "true","test"),
# metrics = c("perc_agr", "bt_user", "ew_user", "ttt_user", "wh_user"),
# suffix = "-user", scales = "free_x")
# plot_by_structure(metric_results_long, "random-forest", origins = c("all", "true","test"),
# metrics = c("perc_agr", "bt_prod", "ew_prod", "ttt_prod", "wh_prod"),
# suffix = "-producer", scales = "free_x")
mle_train_test_iters <- metric_results_long %>%
filter(sample_origin %in% c("true", "test"),
#sample_fraction %in% c("all-data", "67-33", "5-fold"),
metric == "perc_agr",
iter_n %in% c(1,4,13,19),
model == "max-likelihood",
metric != "purity") %>%
ggplot(., aes(y = value)) +
geom_violin(aes(x = sample_structure, fill = sample_fraction), scale = "area", draw_quantiles = c(0.05,0.5,0.9), lwd=0.25) +
scale_fill_manual("Resampling design", values = c("#969696", "#d55e00", "#f0e442", "#56b4e9")) +
#scale_colour_manual("Sample type", values = c("#969696", "#fdae6b", "#d94801")) +
ylab("Metric value") + xlab("Stratification design") +
theme_bw() +
facet_grid(iter_n ~ sample_origin, scales = "free_x", space = "free", drop = T)
ggsave("plots/max-likelihood_iters.png", plot = mle_train_test_iters, device = "png", width = 20, height = 13)
mle_iter_cummean <- metric_results_long %>%
filter(iter_n == 1, model == "max-likelihood", sample_structure != "all-data", metric == "perc_agr", sample_origin == "test") %>%
select(sample_structure, sample_fraction, value) %>%
group_by(sample_structure, sample_fraction) %>%
mutate(iterations = 1:n(), cum_med = cummean(value), scenario = paste0(sample_structure,"_",sample_fraction)) %>%
ggplot(., aes(y = cum_med, x = iterations)) +
geom_line(aes(colour = scenario)) +
#scale_colour_brewer(palette = "RdYlGn") +
ylab("Overall Accuracy") + xlab("Number of iterations") + theme_bw() +
facet_wrap(facets = ~sample_fraction)
ggsave("plots/mle_perc-agr_efficiency.png", plot = mle_iter_cummean, device = "png", width = 12, height = 8)
# readRDS("metric_results_long_reps.rds")
# metric_results_long_reps$reps <- factor(metric_results_long_reps$reps,
# levels = c("5-reps","50-reps","100-reps"))
# metric_results_long_reps$sample_origin <- factor(as.character(metric_results_long_reps$sample_origin),
# levels = c("true", "train", "test", "all"))
# mle_percagr_reps <- metric_results_long_reps %>%
# filter(sample_origin %in% c("all", "train", "test"),
# #sample_fraction %in% c("all-data", "67-33", "5-fold"),
# model == "max-likelihood",
# metric == "perc_agr") %>%
# ggplot(., aes(y = value)) +
# geom_boxplot(aes(x = sample_structure, fill = sample_fraction), outlier.size = 0.25, lwd=0.25) +
# scale_fill_manual("Resampling design", values = c("#969696", "#969696", "#cb181d", "#fc9272", "#31a354")) +
# #scale_colour_manual("Sample type", values = c("#969696", "#fdae6b", "#d94801")) +
# ylab("Metric value") + xlab("Stratification design") +
# theme_bw() +
# facet_grid(reps ~ sample_origin, scales = "free", space = "free", drop = T)
# ggsave("plots/mle_perc-agr_efficiency.pdf", plot = mle_percagr_reps, device = "pdf", width = 20, height = 13)
# junk --------------------------------------------------------------------
# # plot everything... yuk!
# big_plt <- ggplot(data = metric_results_long, aes(y = value)) +
# geom_boxplot(aes(x = sample_structure, fill = sample_origin, colour = sample_fraction), outlier.size = 0.25, lwd=0.25, notch = F) +
# scale_colour_manual(values = c("#969696", "#969696", "#cb181d", "#fc9272", "#31a354")) +
# scale_fill_manual(values = c("#969696", "#9e9ac8", "#fdae6b", "#d94801")) +
# theme_classic() +
# facet_wrap(~ model + metric, ncol = 5, scales = "free")
# ggsave("plots/method_metric_facet.pdf", plot = big_plt, device = "pdf", width = 20, height = 10)
# plot max-like, group by sample design
# mle_train_test1 <- metric_results_long %>%
# filter(sample_origin %in% c("all", "train", "test"),
# #sample_fraction %in% c("all-data", "67-33", "5-fold"),
# model == "max-likelihood",
# metric %in% metrics) %>%
# ggplot(., aes(y = value)) +
# geom_violin(aes(x = sample_structure, fill = sample_fraction), scale = "area", draw_quantiles = quants, lwd=0.25) +
# scale_fill_manual("Resampling design", values = c("#969696", "#969696", "#cb181d", "#fc9272", "#31a354")) +
# #scale_colour_manual("Sample type", values = c("#969696", "#fdae6b", "#d94801")) +
# ylab("Metric value") + xlab("Stratification design") +
# theme_bw() +
# facet_grid(metric ~ sample_origin, scales = "free", space = "free", drop = T)
# ggsave("plots/mle_train_test1.pdf", plot = mle_train_test1, device = "pdf", width = 20, height = 13)
# # old pa plot
# pa_plot <- ggplot(data = pa_results, aes(y = perc_agr)) +
# geom_boxplot(aes(x = type, colour = scenario, fill = method), notch = T) +
# scale_fill_manual(values = c("#fcbba1", "#fb6a4a", "#d4b9da", "#99d8c9", "#238b45")) +
# scale_colour_manual(values = c("#252525", "#e31a1c", "#3f007d"))
# ggsave("perc-agr_results.pdf", plot = pa_plot, device = "pdf", width = 10, height = 5)
# # test perc agreement on smaller chunks of iters and individual iters
# iter_n_breaks <- list(1:30, 31:60, 61:90, 91:120, 121:150, 151:180, 181:210, 211:240, 241:270, 271:300)
# pa_res_plots <- lapply(X = iter_n_breaks, FUN = plot_pa_results, pa_results)
# ggsave("perc-agr_results_30iters.pdf", plot = grid.arrange(grobs = pa_res_plots, ncol=2),
# device = "pdf", width = 20, height = 10)
#
# iter_n_breaks <- as.list(sample(1:300,10))
# pa_res_plots <- lapply(X = iter_n_breaks, FUN = plot_pa_results, pa_results)
# ggsave("perc-agr_results_indiv-iter.pdf", plot = grid.arrange(grobs = pa_res_plots, ncol=2),
# device = "pdf", width = 20, height = 10)