PERFORMANCE ANALYSIS OF THE PROPOSED MECHANISM

PERFORMANCE ANALYSIS OF THE PROPOSED MECHANISM

RESULTS AND DISCUSSION

5.1 INTRODUCTION

The performance analysis of the proposed system shown in this section. Air pollution can defined as the introduction of harmful substances known as pollutants into the earth’s atmosphere. Air pollution is a significant threat to life and the ecosystem as a smart environment by developing a user-configurable air pollution monitoring device through the application of the Internet of Things (IoT). Internet of Things platform, which is a user-configurable device monitored the air pollution, collects location coordinates, and prediction of air pollution.  Air pollution causes a significant threat to life and the earth’s environment. The air pollutants into the atmosphere cause air pollution by changing this natural composition and damaging the ecosystem, thereby making it difficult for the living organisms to the survivability of humans. Air pollution is a severe problem in the developing world, which increases with an increase in population. An air quality monitor is a device that measures the level of common air pollutants. Monitors are available for both indoor and outdoor settings. Sensor-based instruments and air quality monitoring systems are used widely in outdoor ambient applications. The air quality monitoring system used to estimate the data, whether good or bad, or satisfactory results according to the predicted air quality results. The proposed Embedded System Based IoT Enabled Air Pollution Monitoring System used to monitor the air quality levels with efficient data storage in the HTTP server through IoT based fog computing approach. The proposed prediction of air pollution system using the Modified Fruitfly Optimization Algorithm and Enhanced Artificial Neural Network approach used. The E-ANN and MFOA enable responsible performance for categorizing normal or abnormal condition of air at a particular location. The performance analyses of the proposed method evaluated efficiently. They compared with existing approaches results such as accuracy, specificity, sensitivity, Recall, and precision values of the air quality systems.

5.2 PERFORMANCE ANALYSIS OF THE PROPOSED MECHANISM

The overall functionality of the system was demonstrated by carrying the experiments in distinctive settings – classroom and an open environment. At the school – (8m*10m), it was assessed that one sensing node was sufficient, placed within the center aspect of the room at the height of 2m for the open atmosphere, at 9m the adequate height of the node kept.  Air pollution monitoring analysis, measurements have been carried within the boy’s hostel building of the K.L.N. College of Information Technology from 1 May to 15 May 2018. We decided 15 consecutive days and 24 readings each day (every hour) to determine the consistency of the measurement. Each monitoring example analyzed various environmental and weather parameters like temperature, relative humidity, and wind speed.

Air monitoring information examination AQI for PM2.5 revealed in figure 5.1. The validation of the trustability and the achievement of the system is to be measured; additional data sets and identified PM2.5 provided other data sets, as shown in figure 5.2. Over to verify the trustworthiness of the system, more information sets were acquired from recognized PM2.5 databases http://app.cpcbccr.com/AQI_India. These datasets are correlated in Figures 5.1 and 5.2 for outdoor and indoor environments, respectively.

Figure 5.1 Indoor measurement of data analysis in PM2.5 AQI value association with official data, for 15 days from 1 May to 15 May 2019.

Figure 5.2 free measurement of data analysis in PM2.5 AQI value association with official data, for 15 days from 1 May to 15 May 2019.

It is evident from Figure 5.3 that the aims of the AQI range accord were well when measured with various sensing nodes as correlated to the single-node illustration. Although the data trends were equivalent to official data in the single node (N=1) scenario, the gap and difference are higher than in the multiple node (N=2) cases. The sensing performance increases with the study of more sensing nodes because it improves the quality of the data by recognizing the sensor drift and improving the coverage capability. Additionally, the multiple node method somehow combats the issue of non-uniform pollution density. The inconsistency between the measurements and reference data may be due to many factors, such as the various measurement locations, respective environment, data acquisition techniques, and the type of the consolidated sensors.

Figure 5.4 shows the AQI trends measured in the indoor environment. Those measurements have a more concise discrepancy as correlated to the outdoor fields; this is because environmental factors have some effects in the closed environment. Consequently, the air quality information collected by this monitoring system was able to determine the accuracy and reliability necessities.

Figure 5.3 Comparison between the measurements with single and multiple sensing nodes from 01-15 May 2019

Figure 5.4 Comparison between the measurement and reference data on total AQI (Indoor) from 01-15 May 2019.

Some of the atmospheric properties like wind speed, temperature, and humidity have a link to PM2.5 concentrations. The effect of these parameters on PM2.5 concentrations observed; Figure 5.5 plots the relationships among PM2.5 and the wind speed. Now, observation says that from 1 May to 15 May the PM 2.5 concentrations improved progressively with the wind speeds touching approximately 3.5 m/s. Subsequently, the speeds decreased slowly and held in a lower range, which was owing to the dispersal of the pollutants by strong winds. It is possible to decide that AQI depends on the wind speed variation to some extent, as PM2.5 is an essential participant of the AQI.

Figure 5.5 Result of the wind speed on PM 2.5 concentration (microgram per cubic meter)

5.3 PERFORMANCE ANALYSIS OF THE PROPOSED MECHANISM

This section provides the analysis of the performance of the proposed scheme with the study area, datasets in the Madurai location, and their performance results.

5.3.1 Study area (Madurai city)

Figure 5.6 represents the view of satellite in Madurai city that was highlighted in 3 monitoring stations and located at 9 ° 54’N, 78° 84’E, and 100 m over sea level. Commonly recognized as Temple City that has a significant cultural, religious, and historical heritage in South India.

Figure 5.6 Satellite picture of Madurai city showing three monitoring sites

The city’s ambient air quality is significantly affected because of the fossil fuel combustion in mobile and stationary sources, and it notified that the combustion rate of fossil fuels had an escalating trend over decades.

5.3.2 Data set Description:

The data set of ambient air quality gathered from the Tamilnadu pollution control board (TNPCB) and central pollution control board (CPCB) ambient air quality data are existing on the websites. This study pertains to the Chennai region Tamilnadu (India) on behalf of the ambient air quality (http://www.tnpcb.gov.in/ ambient_airquality.htm) over nine years (January 2007 to December 2015). The temperature of meteorological data (in °C), relative humidity (in %), wind speed (in MPs), and rain (in mm) have collected from Madurai city’s meteorological section. Table 5.1 depicts the standard deviation, mean, maximum, and minimum values of pollutant concentrations and meteorological parameters for the year 2006-2015. The mean annual temperature is about 30 °C, yearly relative humidity is among 17.15, and 65.91% with the existing direction of wind NNE to NNW and annual speed of the wind is among 2 and 13.08mps. There will be the occurrence of heavy rain in the September to December period, which is between 0 and 70 mm.

Table 5.1 Daily average means, standard deviation, minimum and maximum values of Meteorological parameters and pollutant concentrations of PM₁₀, SO₂, and NO₂

The site characteristics were dissimilar that ensures industrial (Fenner), traffic (KunnathurChatram), and residential (High way building). The monitoring sites of Air quality categorized as traffic, residential and industrial sites, and the monitoring of air pollution carried for the 2006-2015 period. The issues occurring with these data sets were missing data and outliers. The outlier is because of the instrument failure or measurement of the pollutant in an improper manner. The outliers are minimum or maximum values of data. They are estimated with care, as they cause more variation in the model prediction and development. The data Lost is because of the calibration of the instrument or the breakdowns with this the problem was extremely restricted (2%), these gaps are handled/resolved by the method of linear interpolation. To maintain the neural network to process the data efficiently, the entire input variables normalized to the (0,1) range with Eq. 1.

(6)

X_n is a data that were normalised,

X_max maximum value of the measured data

X_i actual measured data,

X_min minimum amount of the measured data,

On the whole, the average concentrations of annual PM₁₀ are beyond the national ambient air quality standard (NAAQS) at Fenner and KunnathurChatram. They are nearly at the stage of alarming in the Highway building. The concentrations of PM₁₀ mostly caused by industrial activity, adverse road conditions, and construction activity. The annual average NO₂ and SO₂ levels are under standards NAAQS. The NO₂ levels elevated in an industrial area (Fenner) and traffic sites (kunnathurChatram), and it confirms industries and traffic to be the most critical NO₂ sources in Madurai city. The concentration of SO₂ is very low (insignificant), which is mostly because of diesel vehicles and old vehicles, commercial burning, and industrial burning of different fuel oils.

The neural network optimization is the most significant intention for the development of ANN dependent approaches. The optimization process shows a substantial part in the performance and selection of the network. Therefore, a method of optimization conceded with several MSE and neurons. After that, the multilayer neural network estimated with the use of the BP algorithm with nodes 10, 12, and 18 in the hidden layer. By the increment in neuron number, the several local minimum values provided by the network and various amounts of MSE for the training set and escalating neuron number to the excess of 20 offers results that were unrealistic for the entire pollutants. Table 5.2 shows the performance indices for the testing data for three monitoring sites

Table 5.2 Performance indices for the testing data for three monitoring sites

Figure 5.7 shows the observed and predicted PM₁₀, NO₂ and SO₂ concentrations using developed ANN models.

Figure 5.7 Measured and predicted concentrations of PM₁₀, SO₂ and NO₂ for training data

5.3.3 Performance analysis of the proposed mechanism

The performance analysis of the proposed technique is estimated, and the outcomes are compared to that of the existing methodologies to prove the effectiveness of the proposed scheme. The performance results evaluated in terms of accuracy, sensitivity, specificity, precision, and recall. The proposed E-ANN algorithm compared with existing classifiers like SVM, and Random forest classifier (RF) to estimate the efficacy of this proposed system.

Figure 5.8 Comparative analysis of Accuracy

Figure 5.8 shows the comparative analysis of Accuracy. Figure 5.9 represents the comparative analysis of sensitivity.

Figure 5.9 comparative analysis of sensitivity

Figure 5.10 comparative analysis of specificity

Figure 5.10 represents the comparative analysis of specificity.

Figure 5.11 comparative analysis of precision

Figure 5.12 comparative analysis of Recall

Figure 5.11 represents the comparative analysis of precision. Figure 5.12 illustrates the comparative analysis of Recall. From the analysis, it was evident that the proposed methodology is better in providing enhanced accuracy, sensitivity, specificity, precision, and recall on comparing other existing approaches.