Desktop version

Home arrow Computer Science

  • Increase font
  • Decrease font


<<   CONTENTS   >>

EXPERIMENTAL ANALYSIS

Table 3.1 provides a detailed dataset description of the presented model against three datasets, respectively. The GPS trajectories dataset includes a total of 163 instances under two class labels. Besides, the indoor user movement prediction dataset comprises a set of 13,197 instances, whereas 527 instances are present in water treatment plant.

The performance measures are used to examine throughput, energy consumption, sensitivity, specificity, and accuracy.

FS Results Analysis

Table 3.2 provides the analysis of the FS process of the BOAFS-GBT model in terms of number of features and optimized features.

On the applied dataset, the BOAFS-GBT model has chosen a set of 6 features out of 15 features from GPS trajectories dataset, 2 features from a set of indoor user movement prediction dataset, and 16 features from water treatment plant dataset.

Classification Results Analysis

Table 3.3 and Figures 3.3-3.5 perform the analysis of the classifier models of the proposed BOAFS-GBT model in terms of distinct performance measures. On the applied GPS trajectories dataset, the NBTree model has attained better results with the minimum sensitivity of 87.09%, specificity of 87.23%, and accuracy of 88.36%. At the same time, the MLP model has shown extraordinary results by achieving slightly higher sensitivity of 92.40%, specificity of 89.10%, and accuracy of 90.09%. Likewise, the SVM model has shown somewhat

TABLE 3.1 Dataset Description

Dataset

No. of Instances

Number of Classes

GPS trajectories

163

2

Indoor user movement prediction

13,197

2

Water treatment plant

527

2

TAB LE 3.2 Results Analysis of Feature Selection

Dataset

No. of Features

Optimized Features

GPS trajectories

15

6

Indoor user movement prediction

4

2

Water treatment plant

38

16

TABLE 3.3 Results Analysis of Classification Performance

Methods

GPS Trajectories

Movement Prediction

Water Treatment

Sensitivity

Specificity

Accuracy

Sensitivity

Specificity

Accuracy

Sensitivity'

Specificity

Accuracy

BOAFS-GBT

96.78

97.10

97.88

95.83

96.92

96.87

94.20

94.82

94.88

GBT

95.88

96.23

95.89

93.58

92.64

92.65

92.49

90.79

90.63

SVM

91.37

90.44

91.35

90.98

89.05

90.68

90.48

89.05

87.57

MLP

92.40

89.10

90.09

89.57

88.47

89.07

89.46

89.57

88.66

NBTree

87.09

87.23

88.36

85.48

84.20

95.35

82.93

80.58

81.59

Classification analysis of BOAFS-GBT model on GPS trajectories dataset

FIGURE 3.3 Classification analysis of BOAFS-GBT model on GPS trajectories dataset.

higher classifier outcome with the sensitivity of 91.37%, specificity of 90.44%, and accuracy of 91.35%. Likewise, the GBT model has shown moderate performance with the sensitivity of 95.88%, specificity of 96.23%, and accuracy of 95.89%. Finally, the BOAFS-GBT model has shown superior results with the maximum sensitivity of 96.78%, specificity of 97.10%, and accuracy of 97.88%.

On the applied movement prediction dataset, the NBTree method has reached considerable results with the lower sensitivity of 85.48%, specificity of 84.20%, and accuracy of 95.35%. Simultaneously, the MLP method has showcased tremendous outcome by achieving better sensitivity of 89.57%, specificity of 88.47%, and accuracy of 89.07%. Similarly, the SVM approach has illustrated manageable classifier outcome with the sensitivity of

Classification analysis of BOAFS-GBT model on movement prediction dataset.

FIGURE 3.4

Classification analysis of BOAFS-GBT model on water treatment dataset

FIGURE 3.5 Classification analysis of BOAFS-GBT model on water treatment dataset.

90.98%, specificity of 89.05%, and accuracy of 90.68%. Along with that, the GBT technology has demonstrated gradual function with the sensitivity of93.58%, specificity of92.64%, and accuracy of 92.65%. Finally, the BOAFS-GBT approach has depicted supreme outcome with the higher sensitivity of 95.83%, specificity of 96.92%, and accuracy of 96.87%.

Similarly, on the applied water treatment dataset, the NBTree technique has accomplished better outcomes with the lower sensitivity of 82.93%, specificity of 80.58%, and accuracy of 81.59%. The MLP method has illustrated remarkable outcome by attaining moderate sensitivity of 89.46%, specificity of 89.57%, and accuracy of 88.66%. Similarly, the SVM approach has depicted gradual classifier results with the sensitivity of 90.48%, specificity of 89.05%, and accuracy of 87.57%. Likewise, the GBT approach has demonstrated reasonable function with the sensitivity of 92.49%, specificity of 90.79%, and accuracy of 90.63%. Consequently, the BOAFS-GBT technology has provided best outcomes with the higher sensitivity of 94.20%, specificity of 94.82%, and accuracy of 94.88%.

Energy Consumption Analysis

Table 3.4 and Figure 3.6 provide an energy consumption analysis of the presented model with and without the optimization model. Under the data size of 50, the Fladoop without

TABLE 3.4 Energy Consumption (J) Analysis

Data Size

Hadoop with Optimization

Hadoop without Optimization

50

125

180

100

160

210

150

180

240

200

235

255

250

260

290

300

295

320

Energy consumption analysis with and without the optimization model

FIGURE 3.6 Energy consumption analysis with and without the optimization model.

optimization requires higher energy consumption of 180 J, whereas the Hadoop with optimization necessitate lower energy consumption of 125 J. Initially, under the data size of 100, the Hadoop with no optimization needs maximum power application of 210 J, whereas the Hadoop with optimization requires least energy application of 160 J.

Next, under the data size of 150, the Hadoop that lacks optimization demands for greater energy consumption of 240 J, whereas the Hadoop with optimization needs minimal energy consumption of 180 J. Then, under the data size of 200, the Hadoop with no optimization necessitate greater power utilization of 255 J, whereas the Hadoop with optimization requires minimum power application of 235 J. Afterward, under the data size of 250, the Hadoop without optimization acquires maximum power utilization of 290 J, whereas the Hadoop with optimization requires low power application of 260 J. Simultaneously, under the data size of 300, the Hadoop with no optimization needs maximum power consumption of 320 J, whereas the Hadoop with optimization requires minimum energy consumption of 295 J.

Throughput Analysis

Table 3.5 and Figure 3.7 provide the throughput analysis of the presented model with and without the optimization model. Under the data size of 50, the Hadoop without

TABLE 3.5 Throughput (Kbps) Analysis

Data Size

Hadoop with Optimization

Hadoop without Optimization

50

92

86

100

95

89

150

96

91

200

94

88

250

98

93

300

97

92

Social Internet of Things

47

Throughput analysis with and without the optimization model

FIGURE 3.7 Throughput analysis with and without the optimization model.

optimization achieves minimal throughput of 86 Kbps, whereas the Hadoop with optimization offers higher throughput of 92 Kbps. Under the data size of 100, the Hadoop that lacks optimization accomplishes lower throughput of 89 Kbps, whereas the Hadoop with optimization provides maximum throughput of 95 Kbps. Next, under the data size of 150, the Hadoop with no optimization attains lower throughput of 91 Kbps, whereas the Hadoop with optimization gives better throughput of 96 Kbps. Similarly, under the data size of 200, the Hadoop with no optimization obtains least throughput of 88 Kbps, whereas the Hadoop with optimization produces greater throughput of 94 Kbps.

Likewise, under the data size of 250, the Hadoop with no optimization reaches lower throughput of 93 Kbps, whereas the Hadoop with optimization gives maximum throughput of 98 Kbps. Finally, under the data size of 300, the Hadoop with no optimization attains lower throughput of 92 Kbps, whereas the Hadoop with optimization generates remarkable throughput of 97 Kbps.

 
<<   CONTENTS   >>

Related topics