• Tiada Hasil Ditemukan

Modeling severity of road traffic accident in Nigeria using artificial neural network

N/A
N/A
Protected

Academic year: 2022

Share "Modeling severity of road traffic accident in Nigeria using artificial neural network"

Copied!
8
0
0

Tekspenuh

(1)

https://doi.org/10.17576/jkukm-2019-31(2)-06

(2)

TABLE 1. Descriptive statistics of data

Variable Mean Standard Deviation Kurtosis Minimum Maximum

GDP/Capita $ 679.20 810.72 2.84 93.00 3,221.70

Population 98,271,483.34 41,356,925.11 -0.88 45,137,812.00 185,989,640.00

Registered Cars 2,300,280.70 2,723,218.14 1.89 0.00 10,600,000.00

GSM Subscription 19,664,193.47 42,905,255.92 3.39 0.00 154,380,000.00

Cases 19,658.93 8,092.76 -0.50 8,477.00 37,881.00

Fatality Index 0.34 0.15 -1.13 0.08 0.59

Injury Index 1.26 0.80 0.91 0.39 3.13

1 r q= q

d

(3)

FIGURE 1. Architecture of the developed ANN models GDP/Capita

Population

Registered Vehicles Fatality Index

Injury Index

Accident Cases

(Input Data set 2 only)NGS

Input Layer 4-5 Neurons

One-Hidden-Layer 4-14 Neurons

Output Layer 2 Neurons

norm min

max min

X x x (U L) L x x

= − − +

1

2

obs pre

j

N N

n

2 2

( )

( )

j obs pre

j obs obs

N N N N

(4)

TABLE 2. Performance result of ANN models

Training Testing

Models R2 RMSE (normalized) R2 RMSE (normalized)

Fatality Injury Fatality Injury Fatality Injury Fatality Injury

Model-1 = Without NGS

4 0.9579 0.9710 0.0354 0.033 0.7726 0.4445 0.0384 0.0384

6 0.9429 0.961 0.0412 0.0329 0.5413 0.4739 0.0546 0.0443

8 0.9707 0.9652 0.0295 0.0301 0.2342 0.5585 0.0705 0.042

10 0.9581 0.9839 0.0353 0.0299 0.4221 0.5651 0.0613 0.0286

12 0.9568 0.9602 0.0359 0.0423 0.4088 0.1296 0.062 0.0449

14 0.9652 0.9651 0.0322 0.0290 0.7727 0.5911 0.0384 0.0421

Model-2 = With NGS

4 0.9318 0.9755 0.0451 0.0367 0.2618 0.3471 0.0693 0.0353

6 0.9554 0.9870 0.0364 0.0313 0.2503 0.5231 0.0698 0.0257

8 0.9652 0.9802 0.0321 0.0331 0.1678 0.4692 0.0736 0.0316

10 0.9432 0.9737 0.0411 0.0331 -0.0216 0.4669 0.0815 0.0365

12 0.9661 0.9919 0.0346 0.0281 0.9598 0.6167 0.0148 0.0201

14 0.9559 0.9897 0.0362 0.0328 0.5855 0.4768 0.0519 0.0228

(5)

TABLE 3. Comparison between the regression, and ANN models in testing

Output Parameter Regression Model ANN Model without GSM ANN Model with GSM

R2 RMSE R2 RMSE R2 RMSE

Fatality index 0.1677 0.0736 0.7727 0.0384 0.9598 0.0148

Injury index 0.3730 0.0726 0.5911 0.0421 0.6167 0.0201

TABLE 4. Results of t-test at 5% level of significance

Output Parameter Regression Model ANN Model without GSM ANN Model with GSM t-stat t-critical t-stat t-critical t-stat t-critical Fatality index -0.0851 ±2.0032 0.0594 ±2.0032 0.2576 ±2.0032 Injury index 0.0026 ±2.0032 0.4932 ±2.0032 -0.2388 ±2.0032

FIGURE 2. Comparison of the fatality index 0.9

1

0.5 0.7

0.3

0

1950 1990

Year

1970 2010

1960 1980 2000

Observed gsm no-gsm MLR

2020

Fatality Index (normalized)

0.8

0.4

0.1 0.6

0.2

(6)

FIGURE 3. Comparison of injury index

FIGURE 4. Model comparison for fatality and injury Index (a, b, c are fatality index plots for Model with GSM, without GSM and MLR

model, and d, e, f are injury index plot for Model with GSM, without GSM and MLR model) 1.2

0

1950 1990

Year

1970 2010

1960 1980 2000

injury index gsm no-gsm MLR

2020

Injury Index (Normalized)

1

0.6

0.2 0.8

0.4

1

1 1

1 1

1 1.2 0.8

0.9 0.9

0.8 0.8

0.8 0.4

0.7 0.7

0.3 0.3

0.4 0.4

0.4 0

0.5 0.5

0.1 0.1

0 0

0 0

0 0

0 0 0

0 0

0.4

0.4 0.4 0.4

0.4 0.4

(a)

(d) (e) (f)

(b) (c)

GSM

GSM

NO-GSM

NO-GSM

MLR

MLR 0.8

0.8 0.8 0.8

0.8 0.8

0.2

0.2 0.2 0.2

0.2 0.2

0.6

0.6 0.6 0.6

0.6 0.6

1

1 1 1

1 1

0.6

0.8 0.8

0.4 0.4

0.6 0.6

0.6 0.2

0.6 0.6

0.2 0.2

0.2 0.2

0.2

(7)
(8)

Rujukan

DOKUMEN BERKAITAN

Artificial Neural Networks, optimization and kinetic modeling of amoxicillin degradation in photo-Fenton process using aluminum pillared montmorillonite supported

Due to this, the study suggests another policy to resolve this problem, termed “Comparable Reputation” (Papaioannou, Reputation-based policies that provide the right incentives

Over the next decade, developing countries like Bangladesh will experience an alarming increase in road accidents and this will continue to remain as a serious challenge.

The proposed voltage sag source location estimation using RBFN was tested on CIVANLAR distribution test system and Permas Jaya distribution network and the results

Method of extraction the image acquisition uses suitable image processing algorithms and then makes recognition and classification of healthy or unhealthy leaves

Figure 3.1 Data collection through simulation using ADS 21 Figure 3.2 Artificial neural network implementation in

On the other hand, the modeling works was covers the development of fault detection and isolation using artificial neural network and fuzzy logic.. Chapter 4 Presents the

The selected neural network architecture is the Multilayer Perceptron (MLP) network, which is trained to recognize the peaks. The MLP network is trained with two