In [1]:

# Harbinger Package
# version 1.0.777

source("https://raw.githubusercontent.com/cefet-rj-dal/harbinger/master/jupyter.R")

#loading Harbinger
load_library("daltoolbox") 
load_library("harbinger")

Loading required package: daltoolbox

Registered S3 method overwritten by 'quantmod':
  method            from
  as.zoo.data.frame zoo 


Attaching package: ‘daltoolbox’


The following object is masked from ‘package:base’:

    transform


Loading required package: harbinger

In [2]:

#loading the example database
data(examples_anomalies)

In [3]:

#Using the tt time series
dataset <- examples_anomalies$tt
dataset$event <- factor(dataset$event, labels=c("FALSE", "TRUE"))
head(dataset)

A data.frame: 6 × 2
	serie	event
	<dbl>	<fct>
1	1.0000000	FALSE
2	0.9689124	FALSE
3	0.8775826	FALSE
4	0.7316889	FALSE
5	0.5403023	FALSE
6	0.3153224	FALSE

In [4]:

#ploting the time series
plot_ts(x = 1:length(dataset$serie), y = dataset$serie)

Warning message:
“Removed 1 rows containing missing values (`geom_point()`).”
Warning message:
“Removed 1 row containing missing values (`geom_line()`).”

In [5]:

# data preprocessing
slevels <- levels(dataset$event)

train <- dataset[1:80,]
test <- dataset[-(1:80),]

norm <- minmax()
norm <- fit(norm, train)
train_n <- transform(norm, train)
summary(train_n)

     serie          event   
 Min.   :0.0000   FALSE:76  
 1st Qu.:0.2859   TRUE : 4  
 Median :0.5348             
 Mean   :0.5221             
 3rd Qu.:0.7587             
 Max.   :1.0000

In [6]:

model <- hanc_ml(cla_rf("event", slevels, mtry = 1, ntree=5))

In [7]:

# fitting the model
model <- fit(model, train_n)
detection <- detect(model, train_n)
print(detection |> dplyr::filter(event==TRUE))
# evaluating the training
evaluation <- evaluate(model, detection$event, as.logical(train_n$event))
print(evaluation$confMatrix)

  idx event    type
1  12  TRUE anomaly
2  24  TRUE anomaly
3  38  TRUE anomaly
4  50  TRUE anomaly
          event      
detection TRUE  FALSE
TRUE      4     0    
FALSE     0     76

In [8]:

# ploting training results
  grf <- har_plot(model, train_n$serie, detection, as.logical(train_n$event))
  plot(grf)

In [9]:

# preparing for testing
  test_n <- transform(norm, test)

In [10]:

# evaluating the detections during testing
  detection <- detect(model, test_n)

  print(detection |> dplyr::filter(event==TRUE))

  evaluation <- evaluate(model, detection$event, as.logical(test_n$event))
  print(evaluation$confMatrix)

  idx event    type
1  10  TRUE anomaly
          event      
detection TRUE  FALSE
TRUE      1     0    
FALSE     1     20

In [11]:

# ploting the results during testing
  grf <- har_plot(model, test_n$serie, detection, as.logical(test_n$event))
  plot(grf)

Warning message:
“Removed 1 rows containing missing values (`geom_point()`).”
Warning message:
“Removed 1 row containing missing values (`geom_line()`).”

In [ ]: