Packages
library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionlibrary(ggplot2)
library(umap)
library(seewave)
library(tuneR)
library(phonTools)##
## Attaching package: 'phonTools'## The following objects are masked from 'package:tuneR':
##
## normalize, play## The following object is masked from 'package:seewave':
##
## preemphasis## The following object is masked from 'package:dplyr':
##
## slicelibrary(signal)##
## Attaching package: 'signal'## The following object is masked from 'package:phonTools':
##
## resample## The following object is masked from 'package:seewave':
##
## unwrap## The following object is masked from 'package:dplyr':
##
## filter## The following objects are masked from 'package:stats':
##
## filter, polylibrary(warbleR)## Loading required package: NatureSounds## Loading required package: knitrlibrary(voice)##
## Attaching package: 'voice'## The following object is masked from 'package:signal':
##
## interp## The following object is masked from 'package:seewave':
##
## durationlibrary(randomForest)## randomForest 4.7-1.1## Type rfNews() to see new features/changes/bug fixes.##
## Attaching package: 'randomForest'## The following object is masked from 'package:ggplot2':
##
## margin## The following object is masked from 'package:dplyr':
##
## combinelibrary(datasets)
library(caret)## Loading required package: latticePart 1: Waveform Analysis
read in files and convert into Wave class type
filenames <- list.files("150data", pattern="*.wav", full.names=TRUE)
ldf <- lapply(filenames, tuneR::readWave)Spectrogram of a closed hi-hat.
hat <- ldf[[3]]
seewave::spectro(hat, f=48000, flog=TRUE, flim=c(0,20), tlim=c(0,0.2), main='hat')
### Spectrogram of a snare drum.
snare <- ldf[[115]]
seewave::spectro(snare, f=48000, flog=TRUE, flim=c(0,20), tlim=c(0,0.2), main='snare')
### Spectrogram of a kick drum.
kick <- ldf[[51]]
seewave::spectro(kick, f=48000, flog=TRUE, flim=c(0,20), tlim=c(0,0.2), main='kick')
### above are just three examples. Different samples end up getting very
hard to tell apart. In this case, the snare spectrogram looks very
similar to this kick drum’s spectrogram:
snare2 <- ldf[[101]]
kick2 <- ldf[[52]]
seewave::spectro(snare2, f=48000, flog=TRUE, flim=c(0,20), tlim=c(0,0.2), main='snare2')
seewave::spectro(kick2, f=48000, flog=TRUE, flim=c(0,20), tlim=c(0,0.2), main='kick2')
Compare dominant frequencies over time of three instrument samples
hatDfreq <- data.frame(seewave::dfreq(ldf[[2]], plot=FALSE, tlim=c(0,0.2), flim=c(0,20)))
snareDfreq <- data.frame(seewave::dfreq(ldf[[115]], plot=FALSE, tlim=c(0,0.2), flim=c(0,20)))
kickDfreq <- data.frame(seewave::dfreq(ldf[[51]], plot=FALSE, tlim=c(0,0.2), flim=c(0,20)))
hatDfreq$instrument <- "Hat"
snareDfreq$instrument <- "Snare"
kickDfreq$instrument <- "Kick"
combined_df <- rbind(hatDfreq, snareDfreq, kickDfreq)
ggplot(data = combined_df, aes(x = x, y = y, color = instrument)) +
geom_line() +
scale_y_log10() +
labs(title = "Dominant frequencies Hat, Snare, and Kick") +
xlab("Time (s)") +
ylab("Frequency (kHz)")
# Part 2: Organizing/Tidying Data
go through training/testing data and add each file to a class based on its filename
# Function to trim strings based on substrings
trim_to_instrument <- function(string) {
if (grepl("snare", string, ignore.case = TRUE)) {
return("snare")
} else if (grepl("hihat", string, ignore.case = TRUE)) {
return("hat")
} else if (grepl("kick", string, ignore.case = TRUE)) {
return("kick")
} else {
stop("Error: No match found for instrument in string '", string, "'")
}
}
class_list <- sapply(filenames, trim_to_instrument)
class_list <- factor(class_list)
class_listhand-crafted features
#medians of amplitude envelopes
medians <- sapply(ldf, seewave::M, simplify = TRUE)
#dominant frequencies each wave
Dfreq <- lapply(ldf, seewave::dfreq, plot=FALSE)
maxes <- as.data.frame(Dfreq) %>% select(-starts_with("x"))
maxes <-sapply(maxes, function(x) max(x, na.rm = TRUE), simplify = TRUE)
#spectral flatnesses
specs <- lapply(ldf, spec, plot=FALSE)
sfms <- sapply(specs, sfm, simplify = TRUE)
#mfccs
M1 <- extract_features(filenames,
features = c("mfcc"),
check.mono=FALSE,
windowShift = 20,
numcep = 4)
mfccs <- subset(M1, section_seq_file == 1)
mfccdf <- mfccs[, -c(1:3)]combine into single dataframe
data <- data.frame(medians, maxes, sfms, class_list)
rownames(data) <- NULL
df <- cbind(data, mfccdf)
df## medians maxes sfms class_list mfcc1 mfcc2
## 1 2.689808e-05 4.68750 0.204609636 hat 179.13176 -8.4186252
## 2 1.146886e-05 7.78125 0.332010126 hat 183.18767 -11.4214664
## 3 5.199873e-07 16.78125 0.374170458 hat 172.16686 -23.3354828
## 4 1.891459e-06 9.37500 0.401794313 hat 178.16386 -14.8884762
## 5 1.624056e-06 13.59375 0.316058994 hat 168.37095 -21.5579093
## 6 1.441962e-06 11.25000 0.350822215 hat 179.17786 -16.5245074
## 7 1.634273e-05 11.62500 0.270590288 hat 177.23093 -16.6687942
## 8 8.952057e-06 9.28125 0.208957330 hat 179.00688 -14.6181504
## 9 1.556757e-05 15.75000 0.371195821 hat 163.77835 -28.3969917
## 10 5.354960e-06 8.06250 0.613808716 hat 185.83359 -14.4250497
## 11 1.473575e-06 9.75000 0.377004882 hat 179.04478 -20.7076699
## 12 1.504905e-06 12.00000 0.489114714 hat 172.61765 -22.0449068
## 13 1.852330e-06 11.71875 0.511710975 hat 173.98228 -19.9100880
## 14 3.831130e-06 8.06250 0.389379558 hat 178.12046 -14.8544594
## 15 6.757226e-06 13.03125 0.548991098 hat 183.30619 -13.9183048
## 16 3.147135e-05 9.93750 0.633846896 hat 184.36135 -15.1086920
## 17 4.252192e-06 14.06250 0.524943734 hat 183.94280 -13.2929628
## 18 1.119627e-07 0.75000 0.360201324 hat 180.19873 -5.3928232
## 19 8.157962e-06 9.00000 0.462939010 hat 184.73543 -16.0771623
## 20 2.937231e-05 5.15625 0.384998770 hat 173.62530 -6.8263906
## 21 6.021016e-06 7.31250 0.337686742 hat 178.45422 -9.7380050
## 22 1.335820e-07 11.34375 0.254739940 hat 182.43873 -14.0990301
## 23 9.898379e-06 11.43750 0.603516666 hat 173.14518 -24.5168624
## 24 1.351084e-06 14.25000 0.233537795 hat 175.43558 -19.9303916
## 25 5.634163e-07 11.43750 0.245426113 hat 161.05774 -31.5444937
## 26 1.192777e-04 1.96875 0.510714357 hat 182.53586 -18.5400257
## 27 8.861115e-06 6.75000 0.564857756 hat 187.79381 -6.8814404
## 28 5.223584e-06 7.31250 0.190131747 hat 178.67598 -9.9941976
## 29 3.365414e-07 8.53125 0.368618928 hat 170.62732 -23.5899544
## 30 1.033127e-06 12.28125 0.455991279 hat 170.09807 -14.4813754
## 31 1.082676e-04 2.62500 0.078853930 hat 179.69730 1.5428021
## 32 1.920318e-07 7.21875 0.249662105 hat 167.60414 -25.1434222
## 33 7.136607e-07 9.09375 0.306253019 hat 184.50257 -9.6611125
## 34 3.691089e-05 4.59375 0.085779730 hat 181.87840 -9.1703514
## 35 2.888238e-06 12.18750 0.474958645 hat 180.49182 -16.3576595
## 36 3.720352e-04 7.96875 0.114538539 hat 164.00627 -25.9433411
## 37 2.566214e-06 0.84375 0.453368444 hat 186.29589 -4.4846892
## 38 7.932772e-06 0.84375 0.548146300 hat 185.23941 -4.7672963
## 39 1.683517e-04 14.25000 0.627535305 hat 174.46764 -20.3421474
## 40 4.926012e-06 7.31250 0.495003509 hat 177.23701 -9.9900238
## 41 1.440345e-06 6.18750 0.500357828 hat 180.22335 -11.8839351
## 42 1.117089e-04 10.40625 0.418998740 hat 178.47703 -20.1238613
## 43 1.473671e-06 10.68750 0.255596307 hat 171.60734 -21.0752801
## 44 5.685891e-07 10.31250 0.229650011 hat 168.86423 -19.4409817
## 45 8.737135e-06 9.84375 0.592358444 hat 182.38106 -15.8901450
## 46 5.052177e-05 11.15625 0.612433193 hat 183.01935 -14.2381211
## 47 2.409363e-05 7.21875 0.470644207 hat 183.84542 -16.8333737
## 48 2.434209e-06 6.09375 0.351552237 hat 185.30403 -13.8566946
## 49 1.560678e-06 0.65625 0.589903132 hat 185.64166 -9.9161559
## 50 5.395438e-06 3.65625 0.601378268 hat 182.68407 -17.8365918
## 51 8.407019e-05 0.09375 0.057142562 kick 161.51651 4.7673882
## 52 5.291000e-04 0.18750 0.024344738 kick 142.53651 17.0964511
## 53 4.457660e-05 0.46875 0.026614052 kick 141.56901 20.3901404
## 54 1.326085e-04 0.18750 0.291048613 kick 152.85266 19.0358667
## 55 5.237076e-05 0.28125 0.125885813 kick 175.25998 3.7429051
## 56 1.055505e-04 0.09375 0.080569046 kick 157.75299 8.9947794
## 57 5.184660e-05 0.09375 0.172993821 kick 166.94362 9.7833854
## 58 2.577285e-04 0.18750 0.060536994 kick 155.66043 13.7528528
## 59 2.152496e-04 0.18750 0.030535781 kick 157.58159 11.6234210
## 60 9.166327e-05 0.09375 0.370810526 kick 170.45926 16.1937310
## 61 1.344033e-05 0.56250 0.131962211 kick 177.00501 -2.5516652
## 62 1.762508e-04 0.18750 0.022893848 kick 151.44987 12.8886139
## 63 1.713291e-05 20.53125 0.103452213 kick 151.12373 26.9114396
## 64 4.193571e-05 0.18750 0.003782977 kick 136.63747 33.7044183
## 65 5.277706e-04 0.46875 0.054743096 kick 167.18728 4.4635062
## 66 2.571393e-04 0.18750 0.094357419 kick 170.93574 8.1687742
## 67 1.294538e-04 0.09375 0.014036870 kick 142.85994 16.1353831
## 68 4.148206e-04 0.09375 0.018002310 kick 126.32164 20.0518934
## 69 8.107040e-05 0.09375 0.011346204 kick 147.76110 15.1611760
## 70 3.817227e-05 0.18750 0.012162666 kick 134.47140 27.9742188
## 71 9.052531e-06 0.18750 0.062030514 kick 164.88101 10.5918608
## 72 4.418969e-05 0.28125 0.009187091 kick 146.68970 29.4229147
## 73 2.320087e-02 0.18750 0.203656220 kick 173.81105 12.8734649
## 74 1.251280e-04 0.18750 0.237988990 kick 158.97455 14.4092694
## 75 3.055273e-02 0.93750 0.007411004 kick 154.68397 20.1766675
## 76 1.369901e-04 0.09375 0.097692966 kick 159.01601 11.4967982
## 77 7.368968e-06 0.37500 0.012102547 kick 136.17771 27.7117063
## 78 7.730536e-05 0.56250 0.030288648 kick 177.43632 6.2760312
## 79 2.239873e-06 18.28125 0.020515445 kick 155.23720 21.1993482
## 80 2.383305e-04 0.09375 0.093970147 kick 155.99331 15.1787798
## 81 4.219157e-05 18.18750 0.027821534 kick 145.87191 21.4673245
## 82 1.175613e-02 0.18750 0.034373083 kick 151.13485 16.0076403
## 83 1.113747e-04 0.18750 0.001355992 kick 95.88751 35.2781636
## 84 1.823330e-04 0.09375 0.097934232 kick 142.05526 11.7680383
## 85 2.002760e-04 0.09375 0.136842276 kick 142.00318 12.8765586
## 86 3.431754e-05 0.28125 0.006041619 kick 134.52313 31.5292740
## 87 3.692497e-04 17.25000 0.276856330 kick 139.97684 21.8515364
## 88 2.350660e-05 0.18750 0.032402699 kick 158.35052 4.0577390
## 89 1.122703e-05 0.18750 0.049876647 kick 162.08072 14.1314438
## 90 1.020991e-05 0.18750 0.172297825 kick 168.86059 9.0104140
## 91 1.562879e-04 0.00000 0.005905879 kick 140.07089 25.7198919
## 92 1.152145e-04 0.28125 0.072869592 kick 161.36616 7.6372015
## 93 9.101402e-05 0.18750 0.051601038 kick 159.12711 7.1974033
## 94 1.727802e-02 22.03125 0.111895926 kick 149.22000 2.9568758
## 95 1.823137e-02 0.28125 0.438407337 kick 165.13194 11.3653763
## 96 2.333464e-04 0.09375 0.111866619 kick 166.53563 7.2131403
## 97 6.769791e-06 5.53125 0.075174343 kick 147.87955 -1.4661615
## 98 5.074300e-05 0.09375 0.049103251 kick 149.01507 18.5185363
## 99 5.279081e-04 0.28125 0.441115736 kick 165.51721 16.0537789
## 100 1.400839e-06 0.46875 0.080649764 kick 172.23041 13.5007293
## 101 1.528673e-05 0.18750 0.103848503 snare 172.91244 -0.4095897
## 102 2.540404e-05 8.43750 0.202207793 snare 174.17290 7.6647496
## 103 7.108802e-06 0.18750 0.311904275 snare 170.43778 9.4053058
## 104 4.334505e-05 4.03125 0.224389603 snare 187.73820 -2.3863878
## 105 8.672130e-06 0.28125 0.562496755 snare 182.70515 -0.5330093
## 106 3.203001e-05 0.18750 0.165788719 snare 180.29698 5.4070671
## 107 3.363245e-05 6.65625 0.166583540 snare 177.45030 5.8556149
## 108 2.985106e-05 1.40625 0.214977059 snare 184.35739 7.0425564
## 109 4.765681e-05 0.56250 0.533183656 snare 191.13263 -0.4223163
## 110 2.702805e-05 3.56250 0.212092730 snare 178.34488 5.0190441
## 111 6.680592e-05 1.40625 0.204702160 snare 176.87073 6.3482302
## 112 5.905738e-04 4.21875 0.250768090 snare 167.34591 11.9822175
## 113 8.817453e-05 0.18750 0.438753213 snare 180.91046 -4.6364954
## 114 4.455403e-04 1.03125 0.222594792 snare 194.51535 1.0809124
## 115 2.514580e-05 16.31250 0.372380482 snare 193.01986 -5.0952279
## 116 8.199202e-05 0.37500 0.167899940 snare 181.73177 3.9410395
## 117 5.972964e-07 1.78125 0.152938480 snare 174.63080 4.7147041
## 118 5.816549e-05 12.65625 0.505344186 snare 192.19724 -2.2318089
## 119 3.019170e-06 1.87500 0.392507857 snare 186.59025 -0.1826460
## 120 1.466286e-03 2.25000 0.135961134 snare 187.34641 2.0747612
## 121 1.393724e-05 17.81250 0.390488101 snare 173.20535 -2.7609877
## 122 3.465540e-06 1.03125 0.086901874 snare 178.05244 14.1895037
## 123 6.545849e-06 1.21875 0.105498052 snare 176.04196 12.1244793
## 124 1.218990e-05 19.21875 0.663095839 snare 183.89248 -1.4887017
## 125 2.418153e-04 6.46875 0.303779875 snare 176.82878 -15.0344748
## 126 2.177171e-06 0.37500 0.261860384 snare 170.99443 8.7250004
## 127 5.607254e-05 4.31250 0.185445415 snare 186.32630 4.7875696
## 128 6.806349e-05 0.28125 0.308343195 snare 173.95145 7.4783151
## 129 2.581208e-05 0.28125 0.354685651 snare 169.83100 6.3200342
## 130 6.062868e-06 3.28125 0.270300918 snare 162.53387 4.5514542
## 131 1.125138e-05 0.28125 0.345136805 snare 176.29271 6.1878243
## 132 1.356422e-04 5.34375 0.148947970 snare 183.04525 7.5239666
## 133 2.407483e-04 0.09375 0.427849714 snare 183.30619 -11.4563491
## 134 2.909831e-05 10.78125 0.596531542 snare 179.05788 -15.1999677
## 135 1.539608e-04 0.18750 0.618347358 snare 179.22177 -17.5669965
## 136 6.455536e-06 8.25000 0.109153746 snare 184.60320 6.7688051
## 137 3.522766e-07 3.93750 0.130465809 snare 172.15718 1.8311727
## 138 4.773910e-05 0.28125 0.133568867 snare 176.52417 0.7682257
## 139 1.145767e-05 2.62500 0.270061477 snare 178.43966 1.5946750
## 140 6.397396e-06 2.62500 0.068455294 snare 183.30186 4.6716860
## 141 2.285158e-05 1.59375 0.200413753 snare 182.35789 3.3827218
## 142 2.157369e-05 0.37500 0.241202598 snare 179.05752 0.7194225
## 143 2.009124e-05 0.56250 0.234461463 snare 161.32226 1.2427679
## 144 2.358965e-06 10.03125 0.195675454 snare 181.22278 3.2475291
## 145 7.407383e-05 0.65625 0.219563810 snare 189.98599 1.6208675
## 146 3.160992e-06 12.93750 0.317142973 snare 185.49079 2.1532158
## 147 3.596121e-05 22.03125 0.693519539 snare 183.20340 -6.5147809
## 148 4.087562e-05 22.03125 0.747349813 snare 183.13749 -5.3873135
## 149 3.694842e-06 0.75000 0.136718674 snare 182.85419 3.8994445
## 150 1.094794e-05 0.93750 0.070344086 snare 170.91342 13.7381390
## mfcc3 mfcc4
## 1 -22.2266497 5.4983729
## 2 -10.2173828 10.9199130
## 3 -17.0900428 8.8959059
## 4 -2.4628270 10.6916431
## 5 -5.5933576 -3.3462257
## 6 -9.3727077 7.2097609
## 7 -8.1429035 10.1847197
## 8 -8.8472196 10.1344239
## 9 9.0947728 6.3065136
## 10 -3.6443965 7.1433724
## 11 -11.1722927 5.5566139
## 12 -6.2609100 7.2653590
## 13 -6.2359357 2.9033833
## 14 -21.5949905 5.2475098
## 15 -6.8435270 -3.5483814
## 16 -8.7167288 0.4392875
## 17 -1.4685915 -2.2987207
## 18 -9.5498280 11.8003222
## 19 -6.4742157 7.4232270
## 20 -21.6432298 4.5835707
## 21 -18.5444434 13.4651215
## 22 -4.9210053 7.2452475
## 23 -3.0723502 2.2103164
## 24 -5.9328622 7.9336039
## 25 -11.7994496 1.0893452
## 26 -9.2621990 -1.1619709
## 27 -4.7386791 4.4770220
## 28 -18.1038554 12.8495306
## 29 -15.6154963 12.7553672
## 30 -11.6899067 10.7248676
## 31 -22.9884041 3.6982331
## 32 -11.1222380 15.8549724
## 33 -6.5425083 10.9950550
## 34 -14.4323975 4.3003312
## 35 -12.6062424 -1.3578396
## 36 -13.3572082 14.5540571
## 37 -1.8118386 5.2001825
## 38 -1.2092119 6.1024916
## 39 -6.5343994 7.3590477
## 40 -18.5949217 13.6261572
## 41 -15.6396394 9.1907323
## 42 -8.9015626 3.3612390
## 43 -15.6784021 5.1031629
## 44 -19.9485736 7.5666933
## 45 -3.1764772 3.1697844
## 46 -2.1782362 3.3429362
## 47 -9.0067124 -2.1119588
## 48 -17.3591962 0.6043887
## 49 -4.4612607 4.6746380
## 50 -10.4394127 -3.9981515
## 51 -3.4100344 9.9637937
## 52 -12.6209638 19.6730449
## 53 8.4046262 2.1887415
## 54 1.8373191 10.8962592
## 55 -4.9588128 18.8004488
## 56 -7.9114989 15.3201896
## 57 -3.5620316 12.9844903
## 58 2.6696999 5.1950539
## 59 -1.0481709 15.9054842
## 60 -6.3762451 6.8466676
## 61 2.7616489 2.3393825
## 62 5.5918683 9.8262825
## 63 -2.4033003 6.0380387
## 64 3.1928141 3.8697912
## 65 9.8352587 5.3042973
## 66 -16.1580704 17.6257356
## 67 -4.7792632 18.2758004
## 68 5.2596512 9.8110658
## 69 -15.6189140 19.2764216
## 70 6.6522885 5.2977546
## 71 -11.4104282 17.5374381
## 72 -3.3255250 5.6113734
## 73 -8.8260564 4.8307718
## 74 -10.4904537 10.3356783
## 75 -15.9480287 4.9772580
## 76 -8.6423947 15.2312660
## 77 4.6676404 10.5653823
## 78 -2.6456542 3.5187983
## 79 -3.7716696 10.5790510
## 80 -16.9711362 17.1360156
## 81 -4.5398614 11.2576129
## 82 -2.8477993 10.6798383
## 83 14.7044891 12.2616238
## 84 -0.0698215 17.5824286
## 85 -0.0753951 16.2584388
## 86 1.7760388 6.6894834
## 87 0.5896923 10.1248785
## 88 -4.2453708 16.4714535
## 89 -10.6671218 11.5033529
## 90 -8.2381613 12.7384228
## 91 2.6894392 8.8823248
## 92 -6.3458429 17.8577149
## 93 -6.1912325 14.6654075
## 94 12.2910900 14.3431856
## 95 5.1099907 8.3290034
## 96 -0.2844363 2.9140864
## 97 -24.5061126 24.0664267
## 98 -2.6489804 2.2634454
## 99 -3.0587087 11.1428467
## 100 -17.3217805 21.2009676
## 101 -1.1403637 15.5037088
## 102 -5.1624171 17.1905847
## 103 -6.2059259 9.0416620
## 104 -14.0258292 9.6712709
## 105 2.6782809 5.0003767
## 106 -12.9218803 14.3728291
## 107 -10.3656917 12.5808927
## 108 -15.1725845 -1.3123136
## 109 -2.0416413 4.1797838
## 110 -7.3721832 15.9157013
## 111 -4.0645482 12.7691133
## 112 -7.4337065 9.9801683
## 113 -6.8005308 12.7963674
## 114 -9.8180964 0.3378207
## 115 -6.8233056 6.0154933
## 116 -9.7855686 10.0624400
## 117 -15.2068961 9.4997140
## 118 -9.6178682 -1.8371937
## 119 -2.6327708 1.9417207
## 120 -11.1593874 6.3810069
## 121 -2.9728552 10.4572877
## 122 -19.1392327 8.5744510
## 123 -22.7988934 11.5024686
## 124 -2.3419208 4.8644868
## 125 -7.0812997 9.9788337
## 126 -9.4846551 11.6504252
## 127 -12.1606480 4.3219996
## 128 -0.5336350 7.9173719
## 129 -1.7490994 9.0260530
## 130 -2.0317069 -1.0434656
## 131 -15.2937853 14.1150659
## 132 -16.2975156 1.3036426
## 133 -3.2786682 9.4530662
## 134 -1.2539340 12.6340527
## 135 -1.6881597 12.7991626
## 136 -6.1249054 12.6215970
## 137 -4.0551442 14.7351238
## 138 -5.5837700 14.8450918
## 139 -14.7924421 3.1231936
## 140 -16.5789251 8.3109233
## 141 -5.3577890 2.3254065
## 142 -10.7966357 9.2839551
## 143 -8.4668109 9.7020077
## 144 -3.6318699 8.0684473
## 145 -7.6611927 7.3208005
## 146 -1.3709559 4.9421209
## 147 -9.2769118 8.1330580
## 148 -8.9384303 7.2204205
## 149 -9.7939161 10.0719736
## 150 -22.5597919 15.1649295convert class list into factor variable, check with frequency table, separate into test and training data
df$class_list <- as.factor(df$class_list)
table(df$class_list)##
## hat kick snare
## 50 50 50#set.seed(222)
ind <- sample(2, nrow(df), replace = TRUE, prob = c(0.7, 0.3))
train <- df[ind==1,]
test <- df[ind==2,]Part 3: Training Random Forest Classifier
create the random forest using the R randomForest package:
rf <- randomForest(class_list~., data=train, proximity=TRUE, ntree=150)
print(rf)##
## Call:
## randomForest(formula = class_list ~ ., data = train, proximity = TRUE, ntree = 150)
## Type of random forest: classification
## Number of trees: 150
## No. of variables tried at each split: 2
##
## OOB estimate of error rate: 11.65%
## Confusion matrix:
## hat kick snare class.error
## hat 33 0 2 0.05714286
## kick 0 33 4 0.10810811
## snare 2 4 25 0.19354839Use newly created random forest to classify its own training data. Resulting accuracy should be 100%
p1 <- predict(rf, train)
confusionMatrix(p1, train$ class_list)## Confusion Matrix and Statistics
##
## Reference
## Prediction hat kick snare
## hat 35 0 0
## kick 0 37 0
## snare 0 0 31
##
## Overall Statistics
##
## Accuracy : 1
## 95% CI : (0.9648, 1)
## No Information Rate : 0.3592
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 1
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: hat Class: kick Class: snare
## Sensitivity 1.0000 1.0000 1.000
## Specificity 1.0000 1.0000 1.000
## Pos Pred Value 1.0000 1.0000 1.000
## Neg Pred Value 1.0000 1.0000 1.000
## Prevalence 0.3398 0.3592 0.301
## Detection Rate 0.3398 0.3592 0.301
## Detection Prevalence 0.3398 0.3592 0.301
## Balanced Accuracy 1.0000 1.0000 1.000classify test data and generate confusion matrix
p2 <- predict(rf, test)
confusionMatrix(p2, test$ class_list)## Confusion Matrix and Statistics
##
## Reference
## Prediction hat kick snare
## hat 14 0 3
## kick 0 13 2
## snare 1 0 14
##
## Overall Statistics
##
## Accuracy : 0.8723
## 95% CI : (0.7426, 0.9517)
## No Information Rate : 0.4043
## P-Value [Acc > NIR] : 3.959e-11
##
## Kappa : 0.8087
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: hat Class: kick Class: snare
## Sensitivity 0.9333 1.0000 0.7368
## Specificity 0.9062 0.9412 0.9643
## Pos Pred Value 0.8235 0.8667 0.9333
## Neg Pred Value 0.9667 1.0000 0.8437
## Prevalence 0.3191 0.2766 0.4043
## Detection Rate 0.2979 0.2766 0.2979
## Detection Prevalence 0.3617 0.3191 0.3191
## Balanced Accuracy 0.9198 0.9706 0.8506randomForest’s tuneRF function to explore different ntree values
#### played around with this function a little bit, it didn't seem to be very helpful given that the tuning method has no patience for anything besides an decrease in error:
t <- tuneRF(train[,-4], train[,4],
stepFactor = 5,
plot = TRUE,
ntreeTry = 50,
trace = TRUE,
doBest=TRUE,
improve = 0.01)## mtry = 2 OOB error = 11.65%
## Searching left ...
## mtry = 1 OOB error = 11.65%
## 0 0.01
## Searching right ...
## mtry = 7 OOB error = 16.5%
## -0.4166667 0.01
p3 <- predict(t, test)
confusionMatrix(p3, test$ class_list)## Confusion Matrix and Statistics
##
## Reference
## Prediction hat kick snare
## hat 14 0 2
## kick 0 13 1
## snare 1 0 16
##
## Overall Statistics
##
## Accuracy : 0.9149
## 95% CI : (0.7962, 0.9763)
## No Information Rate : 0.4043
## P-Value [Acc > NIR] : 2.917e-13
##
## Kappa : 0.8716
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: hat Class: kick Class: snare
## Sensitivity 0.9333 1.0000 0.8421
## Specificity 0.9375 0.9706 0.9643
## Pos Pred Value 0.8750 0.9286 0.9412
## Neg Pred Value 0.9677 1.0000 0.9000
## Prevalence 0.3191 0.2766 0.4043
## Detection Rate 0.2979 0.2766 0.3404
## Detection Prevalence 0.3404 0.2979 0.3617
## Balanced Accuracy 0.9354 0.9853 0.9032Part 4: Classifier Analysis
graph is pretty self explanatory
hist(treesize(rf),
main = "No. of Nodes for the Trees",
col = "green")
### error rate of random forest
plot(rf)
### mfcc2 is the most important attribute followed by mfcc1
varImpPlot(rf, n.var = 7,main = "Variable Importance")
importance(rf)## MeanDecreaseGini
## medians 5.800406
## maxes 7.771504
## sfms 11.844088
## mfcc1 12.951061
## mfcc2 22.435988
## mfcc3 4.083050
## mfcc4 2.870667Partial dependence plot gives a graphical depiction of the marginal effect of a variable on the class probability (classification)
partialPlot(rf, train, sfms, "kick")
#### if the sfm is less than ~0.3, higher chances of classifying into
kick class.
Plot the scaling coordinates of the proximity matrix from randomForest
MDSplot(rf, train$class_list)