Last update: 10.10.2020
Author: M.Sc André Dantas de Medeiros (Universidade Federal de viçosa). Email: [email protected]
This package uses functions developed from the following packages:
pls, prospectr, dplyr, reshape2, ggplot2, ggpubr, factoextra, FactoMineR, readxl, plyr, caret, e1071. We thank the creators of these useful packages.
Warning:
-
The data structure must be in Dataframe format. Use the following command to transform it into a dataframe:
as.data.frame(). -
The last column must contain the response variable in
factororcharacterformat. -
The package is under development and was made for the specific needs of our laboratory.
- install the latest version from GitHub as follow:
#Install
if(!require(devtools)) install.packages("devtools")
if(!require(nira)) devtools::install_github("admedeiros/nira")
#Load package
library(nira)
if(!require(devtools)) install.packages("devtools")
if(require(nira)) remove.packages("nira")
.rs.restartR()
if(!require(nira)) devtools::install_github("admedeiros/nira")
#Load package
library(nira)
import_csv(): Import and merge CSV NIR files
plotmean_df(): Create graphics with the average of the spectra class.
plotraw_df(): Create graphics with the raw of the spectra class.
center_df(): Center the data on the average.
autoscaling_df(): Autoscaling the data using the Z-score
transformation.
snv_df(): Standard Normal Variate (SNV) on near-infrared (NIR) data.
msc_df(): Multiplicative Scatter Correction (MSC) on near-infrared
(NIR) data.
der_df(): Applying derivatives with Savitzky-Golay smoothing..
pca_df(): Plots a graph of the principal component analysis.
classification_df(): Function to develop classification models
load_classifier_df(): Function to load and apply classifier created
with the classification_df function
prediction_df_df(): Function to apply the classifier to new data
without labels
library(nira)
library(ggpubr) #for ggarrange
# raw spectra
a<-plotraw_df(nir_seed)
# mean spectra per class
b<-plotmean_df(nir_seed)
# center
c<-plotraw_df(center_df(nir_seed))
d<-plotmean_df(center_df(nir_seed))
# Autoscaling
e<-plotraw_df(autoscaling_df(nir_seed))
f<-plotmean_df(autoscaling_df(nir_seed))
# SNV
g<-plotraw_df(snv_df(nir_seed))
h<-plotmean_df(snv_df(nir_seed))
# MSC
i<-plotraw_df(msc_df(nir_seed))
j<-plotmean_df(msc_df(nir_seed))
# 1st derivative with Savitzky-Golay smoothing
k<-plotraw_df(der_SG(nir_seed,1))
l<-plotmean_df(der_SG(nir_seed,1))
# 2nd derivative with Savitzky-Golay smoothing
m<-plotraw_df(der_SG(nir_seed,2))
n<-plotmean_df(der_SG(nir_seed,2))
ggarrange(a, b,c,d,e,f,g,h,i,j,k,l,m,n,
labels = c("Raw spectra", "Mean spectra", "Center", "Center","Autoscaling","Autoscaling","SNV","SNV","MSC", "MSC","1st derivative", "1st derivative", "2nd derivative", "2nd derivative" ),
ncol = 2, nrow = 7)
pca_df(nir_seed,1,2)
data(nir_seed)
classification<-classification_df(nir_seed,metrics = 2)
## Warning in nominalTrainWorkflow(x = x, y = y, wts = weights, info = trainInfo, :
## There were missing values in resampled performance measures.
classification$`Cross-validation`
## Partial Least Squares
##
## 210 samples
## 312 predictors
## 3 classes: 'High_quality', 'Intermediate_quality', 'Low_quality'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold, repeated 3 times)
## Summary of sample sizes: 189, 189, 189, 189, 189, 189, ...
## Resampling results across tuning parameters:
##
## ncomp logLoss AUC prAUC Accuracy Kappa Mean_F1
## 1 1.0659706 0.6503401 0.4292712 0.4650794 0.1976190 NaN
## 2 1.0266903 0.7137188 0.4718326 0.5952381 0.3928571 0.5609619
## 3 0.9235199 0.8267574 0.5867059 0.6285714 0.4428571 0.5957456
## 4 0.8862659 0.8396825 0.6121077 0.6666667 0.5000000 0.6461700
## 5 0.8254675 0.9018141 0.6930179 0.7761905 0.6642857 0.7666811
## 6 0.7982458 0.9240363 0.7366469 0.7952381 0.6928571 0.7891360
## 7 0.7491364 0.9557823 0.7852269 0.8761905 0.8142857 0.8730127
## 8 0.7309549 0.9699546 0.8073060 0.8904762 0.8357143 0.8879721
## 9 0.7056253 0.9797052 0.8254200 0.9015873 0.8523810 0.8999919
## 10 0.6879264 0.9858277 0.8337702 0.9317460 0.8976190 0.9304002
## Mean_Sensitivity Mean_Specificity Mean_Pos_Pred_Value Mean_Neg_Pred_Value
## 0.4650794 0.7325397 NaN 0.7591309
## 0.5952381 0.7976190 0.5715248 0.8422302
## 0.6285714 0.8142857 0.6160619 0.8359405
## 0.6666667 0.8333333 0.6617188 0.8451206
## 0.7761905 0.8880952 0.7884099 0.8958894
## 0.7952381 0.8976190 0.8036552 0.9033291
## 0.8761905 0.9380952 0.8847002 0.9417733
## 0.8904762 0.9452381 0.8981834 0.9484829
## 0.9015873 0.9507937 0.9072046 0.9531868
## 0.9317460 0.9658730 0.9393342 0.9684278
## Mean_Precision Mean_Recall Mean_Detection_Rate Mean_Balanced_Accuracy
## NaN 0.4650794 0.1550265 0.5988095
## 0.5715248 0.5952381 0.1984127 0.6964286
## 0.6160619 0.6285714 0.2095238 0.7214286
## 0.6617188 0.6666667 0.2222222 0.7500000
## 0.7884099 0.7761905 0.2587302 0.8321429
## 0.8036552 0.7952381 0.2650794 0.8464286
## 0.8847002 0.8761905 0.2920635 0.9071429
## 0.8981834 0.8904762 0.2968254 0.9178571
## 0.9072046 0.9015873 0.3005291 0.9261905
## 0.9393342 0.9317460 0.3105820 0.9488095
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was ncomp = 10.
classification$Training_results
## Accuracy Kappa Mean_F1
## 0.9666667 0.9500000 0.9665198
## Mean_Sensitivity Mean_Specificity Mean_Pos_Pred_Value
## 0.9666667 0.9833333 0.9668364
## Mean_Neg_Pred_Value Mean_Precision Mean_Recall
## 0.9835141 0.9668364 0.9666667
## Mean_Detection_Rate Mean_Balanced_Accuracy
## 0.3222222 0.9750000
classification$Testing_results
## Accuracy Kappa Mean_F1
## 0.9777778 0.9666667 0.9777778
## Mean_Sensitivity Mean_Specificity Mean_Pos_Pred_Value
## 0.9777778 0.9888889 0.9777778
## Mean_Neg_Pred_Value Mean_Precision Mean_Recall
## 0.9888889 0.9777778 0.9777778
## Mean_Detection_Rate Mean_Balanced_Accuracy
## 0.3259259 0.9833333
classification$plsplot
classification$`Variable importance`
## kernelpls variable importance
##
## variables are sorted by maximum importance across the classes
## only 20 most important variables shown (out of 312)
##
## High_quality Intermediate_quality Low_quality
## 2499.019759 21.486 86.56 100.00
## 2487.033849 19.938 41.06 65.81
## 2103.637378 6.449 51.50 38.96
## 2095.137688 5.984 49.68 37.92
## 2475.162977 16.837 31.14 47.99
## 2086.706843 5.483 45.98 35.71
## 1362.803133 3.437 45.85 35.52
## 2112.206314 6.868 42.66 32.62
## 2078.343145 4.991 41.36 33.61
## 1902.923499 4.935 38.93 27.27
## 2463.404282 13.069 23.00 38.33
## 2330.546543 4.119 37.15 25.30
## 2120.844894 7.765 34.00 28.34
## 2201.896273 2.391 32.73 26.82
## 2070.046224 4.823 32.71 29.98
## 2451.756782 10.717 19.10 32.30
## 1198.95178 1.572 32.30 16.21
## 2005.983045 2.132 32.03 21.53
## 2211.286184 1.471 31.99 24.44
## 2320.118864 3.536 31.90 19.19
data(nir_seed)
load_classifier_df(nir_seed)
## Confusion Matrix and Statistics
##
##
## previsoes High_quality Intermediate_quality Low_quality
## High_quality 100 1 0
## Intermediate_quality 0 94 3
## Low_quality 0 5 97
##
## Overall Statistics
##
## Accuracy : 0.97
## 95% CI : (0.9438, 0.9862)
## No Information Rate : 0.3333
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.955
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: High_quality Class: Intermediate_quality
## Sensitivity 1.0000 0.9400
## Specificity 0.9950 0.9850
## Pos Pred Value 0.9901 0.9691
## Neg Pred Value 1.0000 0.9704
## Prevalence 0.3333 0.3333
## Detection Rate 0.3333 0.3133
## Detection Prevalence 0.3367 0.3233
## Balanced Accuracy 0.9975 0.9625
## Class: Low_quality
## Sensitivity 0.9700
## Specificity 0.9750
## Pos Pred Value 0.9510
## Neg Pred Value 0.9848
## Prevalence 0.3333
## Detection Rate 0.3233
## Detection Prevalence 0.3400
## Balanced Accuracy 0.9725
data(nir_seed)
pc<-prediction_df(nir_seed[,-313])
pca_df(pc)
