# A hybrid optimization with ensemble learning to ensure VANET network stability based on performance analysis

### Section 1

Section 1 contains the results and discussion of the proposed and implemented methods for improving machine learning with a hybrid optimization strategy to predict mobility in VANET. The execution of the project (HFSA-VANET) is evaluated and compared to that of current method (CRSM-VANET). Delay, Energy consumption, Drop, Throughput, and Fairness index measured values ​​are computed and compared to propose (HFSA-VANET) and existing (CRSM-VANET)29 methods. In addition, the implementation is done through NS2 stimulation and comparing the proposed algorithm with these two platforms, along with the windows 10 PRO computer, total RAM capacity of 10 GB, and processor utilized is Intel core (7M) i3-6100CPU @ 3.70GHz processor. The performance metrics are examined in the next section.

#### Performance metrics

Delays occur while a packet travels from its source to its destination.

$$delay= frac{length}{bandwidth}.$$

(19)

It is the number of packets lost as a result of a rogue node (DoS attack).

$$Drop=frac{Send ;packet-Received ;packet}{Send ; packet}.$$

(20)

The throughput refers to the amount of packet data established across a destination, which corresponds to the overall value of packets created by the sender node within a certain time. The formula is as follows:

$$mathrm{Throughput}hspace{0.17em}=hspace{0.17em}mathrm{received ; dates ; packet }times 8/mathrm{data ; packet ; transmission ; period}.$$

(21)

#### Results obtained through node

The performance metrics of the existing technique and the proposed method is compared in the table below.

The primary goal of the performance metrics is to assess the proposed model’s ability to predict mobility in VANET. According to Table 1, when compared to and examined with existing methodology, the proposed method improves machine learning with a hybrid optimization strategy to predict mobility in VANET is more successful.

The delay, energy consumption, drop, throughput and fairness index of the HFSA-VANET and the CRSM-VANET are compared below.

Suggested technique achieves 99 J, 0.093690, 0.897708 for energy consumption, delay value, and drop value in node 20. Furthermore, the new technique achieves a Throughput of 31.341, which is higher than the prior approach. The proposed technique has a fairness score of 7,000,000, whereas the current method has a value of 8,000,000. For energy consumption, delay value, and drop value in node 60, the proposed approach achieves 47 J, 9.752925, 0.472094. In addition, the new method obtains a Throughput of 31,341, which is greater than the previous method. The suggested strategy has a fairness score of 3.000000, compared to 4.000000 for the present method. The proposed technique achieves 36 J, 10.902826, 0.376633 for energy consumption, delay value, and drop value in node 60. Furthermore, the suggested technique achieves a Throughput of 28.423 compared to 26,749 for the existing method. A fairness index value of 2.000000 for the proposed method vs 4.000000 for the existing method is achieved. For energy consumption, delay value, and drop value in node 80, the suggested approach achieves 11 J, 15.287826, 0.116375. Furthermore, as compared to the previous approach, the proposed strategy obtains a Throughput of 18,197. The proposed approach has a fairness index of 1.000000, whereas the present method has a fairness score of 2.000000. The Figs. 3, 4, 5, 6 and 7 are Delay, Energy Consumption, Drop, Throughput, Fairness Index are obtained through node, respectively.

#### Results obtained through speed

The speed of the proposed technique and the existing techniques are compared in terms of delay, energy consumption, drop, throughput, and fairness index. The measured values ​​are demonstrated in the table below. Table 2 shows the speed values ​​of both existing and proposed techniques.

The speed is compared to the delay shown in Fig. 8, speed vs energy shown in FIG. 9, speed vs drop shown in FIG. 10, speed vs throughput shown in FIG. 11, and speed vs fairness index shown in FIG. 12. The speed is compared to the delay, energy, drop, throughput and fairness index, and the graphical representation is shown below.

In speed 20, the proposed approach achieves 1980 J, 1.873793, 19.954160 in terms of energy consumption, delay value, and drop value. In addition, the new method achieves a Throughput of 150, which is greater than the previous method. The suggested approach has a fairness score of 6,000,000, whereas the present method likewise has a 6,000,000 number. The suggested technique achieves 1880 J, 390.117000, 18.883762 for energy consumption, delay value, and drop value in speed 40. Furthermore, the new approach achieves a Throughput of 35, which is higher than the existing method. The recommended technique has a fairness value of 3.000000, but the current method has a score of 4.000000. In speed 60, the suggested approach achieves 2220 J, 654.169557, 22.597974 in terms of energy consumption, delay, and drop value. In addition, the proposed strategy yields a Throughput of 22 vs 16 for the present method. The suggested technique has a fairness index of 2.000000, whereas the present method has a fairness index of 3.000000. The recommended method achieves 880 J, 1223.026093, 9.309993 for energy consumption, delay value, and drop value in speed 80. Furthermore, the new technique achieves a Throughput of 8 and the existing technique achieves a throughput of 6. The suggested technique has a fairness score of 0.000000, whereas the current method has one of 2.000000. The Figs. 8, 9, 10, 11 and 12 are Delay, Energy Consumption, Drop, Throughput, Fairness Index are obtained through speed, respectively. The Section 2 covers the results obtained through the MATLAB software.

### Section 2

This section covers the experimental results obtained through MATLAB (VERSION 2020a) for evaluating the performance with the NS2 tool. Moreover, we also include an additional parameter to ensure the network lifetime of the proposed model. Therefore, the performance can be proven as highly effective as the existing technique. Here, the performance of the proposed model is evaluated using various machine learning approaches such as ANN-HFSA-VANET, SVM-HFSA-VANET, NB-HFSA-VANET, and DT-HFSA-VANET. Thus, the proposed model results can be compared and proven as more effective than all other existing techniques.

Initially, the proposed model is evaluated with ANN-HFSA-VANET, SVM-HFSA-VANET, NB-HFSA-VANET, and DT-HFSA-VANET separately. The following Figs. 13, 14, 15 and 16 are showing the graphical results of ANN, SVM, NB, and DT, respectively. On the other hand, to show comparison based on the aggregation of various machine learning techniques that the proposed method is comparing with the single graphical results.

#### Parameter analysis of ANN-HFSA-VANET

This section deals with the different types of parameters of ANN-HFSA-VANET and is analyzed in the graph shown in Fig. 13.

The figure mentioned above 13 illustrates the various performance analysis based on ANN-HFSA-VANET where (a) shows that the proposed technique has obtained minimum dropout, (b) shows maximum F1 score has been obtained by using the proposed technique, (c) illustrates that maximum packet delivery ratio has obtained for the ANN-HFSA-VANET, (d) and (e) show that proposed ANN-HFSA-VANET has generated high throughput and minimum delay, respectively.

#### Parameter analysis of Decision Tree (DT)-HFSA-VANET

This section deals with the different types of parameters of the Decision Tree and analyzed in the graphs shown in Fig. 14.

Figure 14, shows the analysis of parameters of DT-HFSA-VANET. (a) shows the minimum drop out ratio of the DT-HFSA-VANET, (b) deals with the maximum score for F1 score of DT-HFSA-VANET and its analysis, (c) shows packet delivery ratio of DT-HFSA- VANET and its values ​​plotted, (d) deals with throughput ratio of DT-HFSA-VANET, (e) deals with end to end delay of DT-HFSA-VANET. Standard parameters are analyzed and plotted in a graph and the values ​​increased at the end of each graph of the parameter.

#### Parameter analysis of Navie Baves (NB)-HFSA-VANET

This section deals with the different types of the parameter of Navie Baves and is analyzed in the graph shown in Fig. 15.

In fig. 15a shows that minimum dropout, (b) shows that maximum F1 score, (c) provides maximum packet delivery ratio, (e) shows that minimum delay, respectively for the proposed NB-HFSA-VANET.

#### Parameter analysis of SVM

This section deals with the different types of a parameter of SVM and is analyzed in the graph shown in Fig. 16.

In fig. 16a dropout ratio has obtained with minimum ratio, (b) F1 score has obtained with maximum score, (c) packet delivery ratio has obtained with maximum, and (e) shows minimum delay.

### Parameters for analyzing different data types

This section deals with the parameters of different data and their analysis. The values ​​are plotted in a graph.

From Fig. 17, the parameter analysis value of data types is examined and plotted in a graph where (a) denotes network life time obtained for every second, (b) deals with energy consumption of data packets used per second, (c) deals with through put ratio of data types and their performance, (d) deals with packet delivery ratio of different types of data performance.