Discriminant analysis is a classification method. It assumes that
different classes generate data based on different Gaussian
distributions. To perform Linear Discriminant Analysis in R we will make
use of the lda function in MASS
package.
library(MASS)
library(caret)
library(phonR)
# loading the data:
Arabic <- read.csv("SpeakingRate24.csv")
# delete a column containing Arabic alphabet to avoid bugs.
Arabic <- subset(Arabic, select = -Arabic_phrase)
# delete spaces
Arabic$CVC <- gsub(' ', '', Arabic$CVC)
Arabic$V <- gsub(' ', '', Arabic$V)
Arabic$Length <- gsub(' ', '', Arabic$Length)
Arabic$f0ons <- as.integer(gsub('--undefined--', NA, Arabic$f0ons))
Arabic$f0mid <- as.integer(gsub('--undefined--', NA, Arabic$f0mid))
Arabic$f0off <- as.integer(gsub('--undefined--', NA, Arabic$f0off))
# Transform all variable types from "character" to factors.
Arabic <- dplyr::mutate_if(Arabic, is.character, as.factor)
# display order of V
Arabic$V = factor(Arabic$V, levels = c('i','a','u','i:','a:','u:','e:','o:'))
# display order of C
Arabic$C <- factor(Arabic$C, levels=c("b","d","g","t","k"))
Arabic$VOTi <- as.integer(Arabic$VOTi)
# Normalisation
Arabic$F1midnor <- with(Arabic, normLobanov(F1mid)) # F1mid normalisation
Arabic$F2midnor <- with(Arabic, normLobanov(F2mid)) # F2mid normalisationmodel = lda(V ~ F1midnor+F2midnor, na.action = na.omit, data=Arabic,CV=TRUE)
prev = model$class
# total percent correct
print(sum(prev==Arabic$V)/nrow(Arabic))## [1] 0.5764583# percent correct for each category of V
selectTabs <-table(prev,Arabic$V)
confusionMatrix(selectTabs)## Confusion Matrix and Statistics
##
##
## prev i a u i: a: u: e: o:
## i 558 89 32 129 3 1 368 0
## a 72 532 155 1 212 6 159 93
## u 71 65 416 2 11 222 10 136
## i: 199 22 0 1048 0 0 122 0
## a: 0 254 4 0 899 0 1 65
## u: 0 0 365 0 0 917 0 280
## e: 300 126 5 20 6 0 540 2
## o: 0 112 223 0 69 54 0 624
##
## Overall Statistics
##
## Accuracy : 0.5765
## 95% CI : (0.5665, 0.5864)
## No Information Rate : 0.125
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.516
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: i Class: a Class: u Class: i: Class: a: Class: u:
## Sensitivity 0.46500 0.44333 0.34667 0.8733 0.74917 0.76417
## Specificity 0.92595 0.91690 0.93845 0.9592 0.96143 0.92321
## Pos Pred Value 0.47288 0.43252 0.44587 0.7534 0.73508 0.58707
## Neg Pred Value 0.92375 0.92019 0.90954 0.9815 0.96407 0.96479
## Prevalence 0.12500 0.12500 0.12500 0.1250 0.12500 0.12500
## Detection Rate 0.05813 0.05542 0.04333 0.1092 0.09365 0.09552
## Detection Prevalence 0.12292 0.12812 0.09719 0.1449 0.12740 0.16271
## Balanced Accuracy 0.69548 0.68012 0.64256 0.9163 0.85530 0.84369
## Class: e: Class: o:
## Sensitivity 0.45000 0.5200
## Specificity 0.94536 0.9455
## Pos Pred Value 0.54054 0.5767
## Neg Pred Value 0.92326 0.9324
## Prevalence 0.12500 0.1250
## Detection Rate 0.05625 0.0650
## Detection Prevalence 0.10406 0.1127
## Balanced Accuracy 0.69768 0.7327###
model = lda(V ~ F1midnor+F2midnor+Duration, na.action = na.omit, data=Arabic,CV=TRUE)
prev = model$class
# total percent correct
print(sum(prev==Arabic$V)/nrow(Arabic))## [1] 0.7135417# percent correct for each category of V
selectTabs <-table(prev,Arabic$V)
confusionMatrix(selectTabs)## Confusion Matrix and Statistics
##
##
## prev i a u i: a: u: e: o:
## i 928 184 34 134 2 1 191 1
## a 96 750 153 1 133 7 62 105
## u 68 95 891 2 1 253 3 120
## i: 62 8 0 1039 0 0 128 0
## a: 0 110 1 0 940 0 16 74
## u: 0 0 82 0 0 867 0 257
## e: 46 14 0 24 26 0 793 1
## o: 0 39 39 0 98 72 7 642
##
## Overall Statistics
##
## Accuracy : 0.7135
## 95% CI : (0.7044, 0.7226)
## No Information Rate : 0.125
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.6726
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: i Class: a Class: u Class: i: Class: a: Class: u:
## Sensitivity 0.77333 0.62500 0.74250 0.8658 0.78333 0.72250
## Specificity 0.93488 0.93369 0.93548 0.9764 0.97607 0.95964
## Pos Pred Value 0.62915 0.57383 0.62177 0.8399 0.82384 0.71891
## Neg Pred Value 0.96652 0.94574 0.96216 0.9807 0.96926 0.96033
## Prevalence 0.12500 0.12500 0.12500 0.1250 0.12500 0.12500
## Detection Rate 0.09667 0.07812 0.09281 0.1082 0.09792 0.09031
## Detection Prevalence 0.15365 0.13615 0.14927 0.1289 0.11885 0.12562
## Balanced Accuracy 0.85411 0.77935 0.83899 0.9211 0.87970 0.84107
## Class: e: Class: o:
## Sensitivity 0.66083 0.53500
## Specificity 0.98679 0.96964
## Pos Pred Value 0.87721 0.71572
## Neg Pred Value 0.95320 0.93588
## Prevalence 0.12500 0.12500
## Detection Rate 0.08260 0.06688
## Detection Prevalence 0.09417 0.09344
## Balanced Accuracy 0.82381 0.75232####
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, na.action = na.omit, data=Arabic,CV=TRUE)
prev = model$class
# total percent correct
print(sum(prev==Arabic$V)/nrow(Arabic))## [1] 0.7007292# percent correct for each category of V
selectTabs <-table(prev,Arabic$V)
confusionMatrix(selectTabs)## Confusion Matrix and Statistics
##
##
## prev i a u i: a: u: e: o:
## i 828 153 23 59 2 1 83 1
## a 61 610 108 1 66 2 41 68
## u 38 67 688 2 0 153 0 62
## i: 45 3 0 1068 0 0 142 0
## a: 3 184 6 0 995 0 9 73
## u: 0 1 207 0 1 933 0 309
## e: 221 70 23 70 36 6 921 3
## o: 4 112 145 0 100 105 4 684
##
## Overall Statistics
##
## Accuracy : 0.7007
## 95% CI : (0.6915, 0.7099)
## No Information Rate : 0.125
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.658
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: i Class: a Class: u Class: i: Class: a: Class: u:
## Sensitivity 0.69000 0.50833 0.57333 0.8900 0.8292 0.77750
## Specificity 0.96167 0.95869 0.96167 0.9774 0.9673 0.93833
## Pos Pred Value 0.72000 0.63741 0.68119 0.8490 0.7835 0.64300
## Neg Pred Value 0.95598 0.93174 0.94040 0.9842 0.9754 0.96724
## Prevalence 0.12500 0.12500 0.12500 0.1250 0.1250 0.12500
## Detection Rate 0.08625 0.06354 0.07167 0.1113 0.1036 0.09719
## Detection Prevalence 0.11979 0.09969 0.10521 0.1310 0.1323 0.15115
## Balanced Accuracy 0.82583 0.73351 0.76750 0.9337 0.8982 0.85792
## Class: e: Class: o:
## Sensitivity 0.76750 0.57000
## Specificity 0.94893 0.94405
## Pos Pred Value 0.68222 0.59272
## Neg Pred Value 0.96618 0.93891
## Prevalence 0.12500 0.12500
## Detection Rate 0.09594 0.07125
## Detection Prevalence 0.14062 0.12021
## Balanced Accuracy 0.85821 0.75702#preparing vowel pairs:
Arabica = Arabic[Arabic$Timbre=='a',]
Arabica$Timbre = droplevels(Arabica$Timbre)
Arabica$V = droplevels(Arabica$V)
Arabici = Arabic[Arabic$Timbre=='i',]
Arabici$Timbre = droplevels(Arabici$Timbre)
Arabici$V = droplevels(Arabici$V)
Arabicu = Arabic[Arabic$Timbre=='u',]
Arabicu$Timbre = droplevels(Arabicu$Timbre)
Arabicu$V = droplevels(Arabicu$V)
Arabice = Arabic[Arabic$V=='e:' | Arabic$V=='i',]
Arabice$Timbre = droplevels(Arabice$Timbre)
Arabice$V = droplevels(Arabice$V)
Arabico = Arabic[Arabic$V=='o:' | Arabic$V=='u',]
Arabico$Timbre = droplevels(Arabico$Timbre)
Arabico$V = droplevels(Arabico$V)#a-a:
model = lda(V ~ F1midnor+F2midnor, data=Arabica,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabica$V)/nrow(Arabica))## [1] 0.7683333#a-a: + duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabica,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabica$V)/nrow(Arabica))## [1] 0.9091667#a-a: + word-v-ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabica,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabica$V)/nrow(Arabica))## [1] 0.8775#i-i:
model = lda(V ~ F1midnor+F2midnor, data=Arabici,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabici$V)/nrow(Arabici))## [1] 0.8604167#i-i: + duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabici,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabici$V)/nrow(Arabici))## [1] 0.9233333#i-i: + word-v-ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabici,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabici$V)/nrow(Arabici))## [1] 0.9354167#u-u:
model = lda(V ~ F1midnor+F2midnor, data=Arabicu,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabicu$V)/nrow(Arabicu))## [1] 0.72#u-u: + duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabicu,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabicu$V)/nrow(Arabicu))## [1] 0.8570833#u-u: + word-v-ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabicu,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabicu$V)/nrow(Arabicu))## [1] 0.80875#i-e:
model = lda(V ~ F1midnor+F2midnor, data=Arabice,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabice$V)/nrow(Arabice))## [1] 0.6279167#i-e: + duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabice,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabice$V)/nrow(Arabice))## [1] 0.8766667#i-e: + word-v-ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabice,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabice$V)/nrow(Arabice))## [1] 0.8491667#u-o:
model = lda(V ~ F1midnor+F2midnor, data=Arabico,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabico$V)/nrow(Arabico))## [1] 0.73875#u-o: + duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabico,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabico$V)/nrow(Arabico))## [1] 0.88375#u-o: + word-v-ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabico,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabico$V)/nrow(Arabico))## [1] 0.82875#a-a: slow
Arabic_a=Arabica[Arabica$Rate=='slow',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.78125#a-a: slow +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.9775#a-a: slow +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.8925#a-a: normal
Arabic_a=Arabica[Arabica$Rate=='normal',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.755#a-a: normal +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.9625#a-a: normal +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.85875#a-a: fast
Arabic_a=Arabica[Arabica$Rate=='fast',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.7725#a-a: fast +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.94625#a-a: fast +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.8775######
#i-i: slow
Arabic_i=Arabici[Arabici$Rate=='slow',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.9175#i-i: slow +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.97125#i-i: slow +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.96625#i-i: normal
Arabic_i=Arabici[Arabici$Rate=='normal',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.8775#i-i: normal +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.96375#i-i: normal +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.9375#i-i: fast
Arabic_i=Arabici[Arabici$Rate=='fast',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.81625#i-i: fast +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.9025#i-i: fast +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.9025##########################
#u-u: slow
Arabic_u=Arabicu[Arabicu$Rate=='slow',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.77875#u-u: slow +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.955#u-u: slow +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.88625#u-u: normal
Arabic_u=Arabicu[Arabicu$Rate=='normal',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.74375#u-u: normal +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.9175#u-u: normal +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.805#u-u: fast
Arabic_u=Arabicu[Arabicu$Rate=='fast',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.655#u-u: fast +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.85125#u-u: fast +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.745################
#i-e: slow
Arabic_e=Arabice[Arabice$Rate=='slow',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.59625#i-e: slow +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.95#i-e: slow +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.89125#i-e: normal
Arabic_e=Arabice[Arabice$Rate=='normal',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.6325#i-e: normal +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.93125#i-e: normal +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.845#i-e: fast
Arabic_e=Arabice[Arabice$Rate=='fast',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.6475#i-e: fast +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.91125#i-e: fast +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.81875######################
#u-o: slow
Arabic_o=Arabico[Arabico$Rate=='slow',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.76125#u-o: slow +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.94625#u-o: slow +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.865#u-o: normal
Arabic_o=Arabico[Arabico$Rate=='normal',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.75375#u-o: normal +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.9225#u-o: normal +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.84#u-o: fast
Arabic_o=Arabico[Arabico$Rate=='fast',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.7325#u-o: fast +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.8875#u-o: fast +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.78875#### extreme situation :
##a-slow-a:-fast
Arabic_a=Arabica[Arabica$Rate=='fast' & Arabica$V=='a:' | Arabica$Rate=='slow' & Arabica$V=='a',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.72875#a-slow-a:-fast +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.79875#a-slow-a:-fast +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_a,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_a$V)/nrow(Arabic_a))## [1] 0.8675#####################
##i-slow-i:-fast
Arabic_i=Arabici[Arabici$Rate=='fast' & Arabici$V=='i:' | Arabici$Rate=='slow' & Arabici$V=='i',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.78375#i-slow-i:-fast +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.81#i-slow-i:-fast +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_i,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_i$V)/nrow(Arabic_i))## [1] 0.8875###################
#u-slow-u:-fast
Arabic_u=Arabicu[Arabicu$Rate=='fast' & Arabicu$V=='u:' | Arabicu$Rate=='slow' & Arabicu$V=='u',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.59625#u-slow-u:-fast +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.66375#u-slow-u:-fast +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_u,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_u$V)/nrow(Arabic_u))## [1] 0.73875###########################
#i-slow-e:-fast
Arabic_e=Arabice[Arabice$Rate=='fast' & Arabice$V=='e:' | Arabice$Rate=='slow' & Arabice$V=='i',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.7175#i-slow-e:-fast +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.77625#i-slow-e:-fast +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_e,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_e$V)/nrow(Arabic_e))## [1] 0.8225###########################
#u-slow-o:-fast
Arabic_o=Arabico[Arabico$Rate=='fast' & Arabico$V=='o:' | Arabico$Rate=='slow' & Arabico$V=='u',]
model = lda(V ~ F1midnor+F2midnor, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.69625#u-slow-o:-fast +duration
model = lda(V ~ F1midnor+F2midnor+Duration, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.74375#u-slow-o:-fast +Word_V_ratio
model = lda(V ~ F1midnor+F2midnor+Word_V_ratio, data=Arabic_o,na.action = na.omit, CV=TRUE)
prev = model$class
print(sum(prev==Arabic_o$V)/nrow(Arabic_o))## [1] 0.7975