Content uploaded by Ajay Satija
Author content
All content in this area was uploaded by Ajay Satija on Jan 15, 2023
Content may be subject to copyright.
Content uploaded by Ajay Satija
Author content
All content in this area was uploaded by Ajay Satija on Jan 15, 2023
Content may be subject to copyright.
Municipal Solid Waste Generation
Forecasting for Faridabad City Located
in Haryana State, India
Dipti Singh and Ajay Satija
Abstract Faridabad is the largest metropolitan city in Haryana State (India). Solid
waste management is one of the biggest environmental issues for the municipal
corporation of Faridabad. The Municipal Corporation of Faridabad seems unable to
manage the solid waste due to highly increased urbanization and lack of planning,
funds, and advanced technology. Hence, various private sector companies and
nongovernment organizations are required to work in this sector to resolve such
issues. For the success of such useful work, proper planning is required. Successful
planning depends on the exact prediction of the amount of solid waste generation.
In this paper an artificial neural network model is applied to predict the quantity of
solid waste generation in Faridabad city.
Keywords Solid waste management (SWM) Artificial neural networks (ANN)
Forecasting Municipal solid waste (MSW)
1 Introduction
Solid waste is the refuse material of daily used items discarded by society. The waste
generated from residential sectors, institutional sectors, commercial sectors, and
various industries is an environmental issue. People sufferfrom health problems and
various atmospheric issues such as foul smell, house flies, cockroaches, or other
insects. Hence solid waste management is extremely essential in society. The
growing population, migration to urban areas for employment, gross domestic pro-
duct (GDP) per capita, i.e., living standard parameters, different housing conditions,
global longitude and latitude, seasonal conditions, and regional environmental laws
Dipti Singh (&)
Gautam Buddha University, Greater Noida, India
e-mail: diptipma@rediffmail.com
Ajay Satija
Inderpratha Engineering College, Ghaziabad, India
e-mail: aajaysatija@rediffmail.com
©Springer Science+Business Media Singapore 2016
M. Pant et al. (eds.), Proceedings of Fifth International Conference on Soft
Computing for Problem Solving, Advances in Intelligent Systems
and Computing 437, DOI 10.1007/978-981-10-0451-3_27
285
are the crucial factors affecting solid waste generation. Optimization techniques, soft
computing techniques as ANN models, ANFIS (adaptive neurofuzzy inference
system) models, expert systems, evolutionary algorithms, etc., have been applied for
SWM. Accuracy of waste management forecasting plays an important role in waste
management strategy. Various ANN models have been applied to forecast solid
waste generation in different cities all over the world. But in India not much work is
done in this direction. In this paper ANN model is applied for forecasting of solid
waste in Faridabad city. Before introducing the proposed model, a few research
papers have been revealed in this direction.
Zade and Noori presented an appropriate ANN model with threshold statistics
technique to forecast the solid waste generation in Mashhad city (Iran) [1]. Abdoli
et al. suggested a new approach of removing data trend by taking the logarithm of
data and creating the stationary condition to predict solid waste generation for the
city of Mashhad for the period 2011–2032 [2]. Batinic et al. proposed an ANN
model to forecast the waste characteristics (organic waste, paper, glass, metal,
plastic, and other waste (output parameters)) in Serbia for the period 2010–2026.
The result showed that organic waste will not increase within the period 2010–2016
but within period 2016–2026 there will be expected change. In 2026, 810,000 tons
of solid waste will probably need to be disposed in a landfill area in Serbia [3].
Shahabi et al. suggested a feedforward multilayer perception ANN model to
forecast waste generation in Saqqez City in Kurdistan Province in northwest Iran.
Weekly time series of generated solid waste have been arranged for the period
2004–2007. The authors suggested the Stop Training Approach to solve the
problem of increasing error while training and testing phase of ANN [4]. Patel and
Meka have proposed the feedforward ANN model to forecast Municipal Solid
Waste for the 98 towns of Gujarat for the next 25 years. First, the normalized solid
waste data is applied to ANN model and it is found that the predicted data vali-
dation follows nearly perfect correlation [5]. Roy et al. proposed a feedforward
ANN model to forecast waste generation in Khulna city (Bangladesh). The authors
have arranged the city’s generated solid waste data weekly with the help of time
series. The best neural network model has been selected on the basis of mean
absolute errors and regression R of training, testing, validation, etc. Further results
of observed and predicted values of solid waste generation have been compared [6].
Falahnezhad and Abdoli proposed a neural network model for solid waste gener-
ation in which the effect of preprocessing data is seen. The authors suggest that the
logarithm of input and output data to time series provides more accurate results [7].
Shamshiry et al. suggested an ANN model to forecast the solid waste generation in
Tourist and Tropical Area Langkawi Island, Malaysia, during 2004–2009. The
authors have applied backpropagation neural network for better results. The results
are compared with multiple regression analysis but ANN predicted results seem
much better [8]. Singh and Satija suggest the concept of incineration process for
electricity generation through solid waste in Mumbai and different cities of Gujrat
State [9]. Pamnani and Meka propose an ANN model for forecasting solid waste
generation for small-scale towns and their neighboring villages of Gujarat State,
286 Dipti Singh and Ajay Satija
India. The authors have validated their results with the help of low value of per-
centage prediction error [10].
2 Materials and Methods
2.1 Study Area Identification and MSW Management Issues
Faridabad is the main industrial center (ranked 9th in Asia) and a highly populated
district in Haryana State established by the Sufisaint, Baba Farid, in 1607 AD. It
became the 12th district of Haryana state on 15 August 1979. Its total area is
742.9 km
2
with a dense population of 1,798,954 people (2011 census). There are 91
sectors proposed under the development plan in Faridabad. The city generates per
year, per capita 135.72 kg of waste. In Faridabad only 20.49 % of the population
lives in rural villages (2011 census). There are about 15,000 small, medium, and
large-scale industries, which includes multinational companies and other ISO cer-
tified companies. But the waste is not properly treated at these sites. The Municipal
Corporation of Faridabad (MCF) was established in 1992. MCF was founded by
earlier municipalities of Faridabad Town/New Industrial Township Old Faridabad,
Ballabhgarh and its 38 revenue villages. The related solid waste management issues
follows: improper short- and long-term strategy to manage solid waste related to
MSW Rules 2000, financial issues, high land cost, lack of technical skill in
municipal sanitation workers, lack of sufficient numbers of waste gathering carts or
advanced vehicles, public unawareness about 4-R principles of environment sci-
ences, sanitation work is not followed on Sundays or on other holidays, etc.
2.2 Case Study and Data
The accurate data plays an important role in waste management strategies. But the
collection of reliable and complete solid waste data is a challenging task. The
generated solid waste data is arranged from the MCF head office (NIT Faridabad)
from January 2010 to December 2014 month-wise. According to MCF officials
waste generation (WG) rate is approximately 300 g per day per capita. But there is
month-wise fluctuation in waste generation in the city due to the rapid growth of the
urban population. In 2014 about 133,871,295 kg waste was collected and disposed
by hook loaders in the landfill of village Bandhwari (Gurgaon).
The following are the types of solid waste generated in the city: (a) domestic
waste, (b) commercial waste, (d) institutional waste, (e) industrial waste, (f) agri-
culture waste, (g) energy renewal plant wastes, (h) inert waste, and (i) public place
waste (streets, roads, parks etc.)
Municipal Solid Waste Generation Forecasting …287
In the Indian calendar one year is divided into six seasons. There were seasonal
variations in solid waste generation per day in Faridabad city in 2014. It is observed
that in the monsoon the waste collected is the lowest while in spring the waste
collected is the highest.
2.3 Artificial Neural Network Model (ANN Model)
Neural network models are similar to models of biological systems. In 1943,
Mclloch and Pitts developed the first computational model in which neurophysi-
ology and mathematical logic approaches got merged.
The following model forms the basis of a neural network. Here {x
1
,x
2
,x
3
,…,
x
n
} are the set of inputs to the artificial neurons. The set {w
1
,w
2
,w
3
,…,w
n
} is the
randomly generated weights’set to the input links (synapses). In the biological
neuron system the neuron receives its input signals, sums up them, and produces an
output. If this sum is greater than a threshold value the input signal passes through
the synapse which may accelerate or retard signal. In the ANN model this accel-
eration or retardation is the reason for creation of the concepts of weights. The
weights are multiplied with inputs and transmitted via links (synapses). The larger
weight synapse generates a strong signal while a weak weight synapse generates a
weak signal. The combined input I(say) is transferred to the soma (nucleus–cell
body of neuron in biological system). I=∑w
i
x
i
,i=1,2,…,n. The final output is
transmitted to a nonlinear filter Ф(Transfer function say), i.e., y=Ф(I).
A general form of activation function is the threshold function. This sum is
equated with the threshold value θ, i.e., y=Ф[∑w
i
x
i
−θ] (Fig. 1).
This threshold function may be in the form of heaviside function, Signum
function, hyperbolic tangent function, or sigmoidal function, i.e., UIðÞ¼ 1
1þeaI. The
parameter αis a slope parameter. The three basic classes of ANN models follows:
w2
x3
w1
xn
x2
Weights
wn
w3
x1
Inputs Summation Unit Threshold Unit
Threshold output
Summation of weighted
inputs
Fig. 1 Simple model of artificial neural network
288 Dipti Singh and Ajay Satija
(a) single layer feedforward network (no hidden layer between input and output),
(b) multilayer feed forward network (one or more hidden layers between input and
output), (c) recurrent networks [11]. The tansig transfer function can be used in
input and hidden layer neurons for summing and nonlinear mapping. The purelin
transfer function can be used in linear input–output relationship between the hidden
layers and output layer. Both algorithms may be summarized as (i) IO = tansig
(IN) (ii) FO = purelin (IO). IN, IO, FO represents input, intermediate outputs, and
final output matrices (in Matlab) [5].
There are monthly variations (fluctuation) in estimated solid waste generation.
Hence a dynamic time series neural network model is proposed for forecasting waste
generation. In Matlab 7.12.0 (R2011a) neural network time series tool (ntstool) is
chosen for prediction. Here, nonlinear autoregressive technique is used in which the
series y(t) is predicted with given d past values of y(t). The predicted series can be
formulated as y(t)=f(y(t−1), y(t−2), …,y(t−d)). The input solid waste data is
arranged within January 2010 to December 2014 month-wise, i.e, now 60 months
solid waste collected data is available for prediction. The solid waste targets (rep-
resenting dynamic data) have been taken as 1 ×60 cell array of 1 ×1 matrices. Now
in training data set 70 %, 15 %, 15 % time steps are used for training, validating, and
testing respectively. The network is adjusted according to the training time steps.
Validation time steps represent the generalization of network. The training stops
when generalization stops improving. Testing time steps shows the performance of
the network. Autoregressive neural network may be adjusted by the selecting
number of hidden neurons in hidden layers and number of time delays. If the
network does not perform well, both can be changed. Levenberg–Marquardt back-
propagation algorithm (trainlm) is used to fit inputs and targets. Multiple times
training will generate different results due to different initial conditions. Mean
squared error (MSE) is defined as the average squared difference between outputs
and targets. Lower values are considered better. A zero value shows no error.
Regression Rvalues show the correlation between outputs and target values. The
value R= 1 implies a close relationship between output and targets. Finally, testing
should be done on large data to decide the network performance.
3 Results and Discussions
The statistical analysis of different ANN models is done. Mean square errors
(MSE), root mean square error (RMSE), mean absolute error (MAE), and regres-
sion Rare measured as performance metrics of ANN models. Different structures of
time series neural network models have been investigated with varying values of
hidden layer neurons. Table 1illustrates the ANN model structures, performance
metrics MSE, RMSE, and regression values Rof training, testing, and validation
phases. From table it can be observed that 1-9-1 is the best ANN model for solid
waste prediction due to minimum mean square error and maximum correlation
Rcompared to other ANN models.
Municipal Solid Waste Generation Forecasting …289
Figure 2illustrates the observed and predicted solid waste generation from
training phase of ANN model with structure (1-9-1).
The other performance metric mean absolute error is defined as
MAE ¼1
nX
n
1
w0wp
Here nis the number of months, w
o
is observed solid waste weight, and w
p
is
predicted solid waste present case. The mean absolute error during training in ANN
structure 1-9-1 is 40.91. Figure 3illustrates the different values of Rshowing the
relationship between observed and predicted values of waste during training, val-
idation, and testing phases of the 1-9-1 ANN model (Fig. 4).
Table 1 Results of training, testing, and validation of ANN models
ANN model
structure
MSE RMSE Regression
Training Testing Validation All
1-4-1 3681.44 60.67 0.821987 0.493287 0.754089 0.73493
1-5-1 4378.51 66.17 0.81917 0.50451 0.871937 0.79502
1-6-1 2289.66 47.85 0.89636 0.25555 0.580650 0.72636
1-7-1 3737.45 61.13 0.85026 0.83483 0.357105 0.72468
1-8-1 3120.48 55.86 0.85857 0.600234 0.276443 0.67733
1-9-1 2377.88 48.76 0.890502 0.705548 0.634741 0.82464
1-10-1 3271.61 57.19 0.86355 0.56686 0.394007 0.71777
1-11-1 3257.51 57.07 0.85672 0.252753 0.754561 0.64432
1-12-1 6166.94 78.52 0.725112 0.49032 0.597521 0.63368
1-13-1 2786.23 52.78 0.89632 0.802407 0.811832 0.78396
1-14-1 3095.32 55.63 0.82764 0.38631 0.60608 0.62935
0 10 20 30 40 50 60
0
50
100
150
200
250
300
350
400
450
500
Number of Months
Solid Waste generated in Metric Ton
Observed MSW
Forcasted MSW By ANN 1-9-1 Model
Fig. 2 The observed and predicted solid waste generation from training phase of ANN model
with structure (1-9-1)
290 Dipti Singh and Ajay Satija
4 Conclusion
The aim of this study is to propose an appropriate model for predicting solid waste
for Faridabad city, Haryana (India). Time series neural network tool (Matlab 7.12.0
(R2011a)) has been used for monthly-based solid waste prediction. Different ANN
structures have been trained and tested by changing the number of neurons in
hidden layers. The ANN structure with 9 hidden layer neurons is selected for its
best performance metrics, i.e., low values of MSE, RMSE and high value of R.
Fig. 3 Overall ANN times series response during training testing and validation
Fig. 4 Regression values of training testing, validating, and ALL for 1-9-1 model
Municipal Solid Waste Generation Forecasting …291
The future scope of this work is that the ANFIS (adaptive neuro-fuzzy inference
system) technique can be used on collected data for increasing accuracy of waste
generation forecasting. Such model may be generalized as solid waste prediction on
weekly basis for its fine prediction.
References
1. Zade, J.G.M., Noori, R.: Prediction of municipal solid waste generation by use of artificial
neural networks: a case study of Mashhad. Int. J. Environ. Res. 2(1), 13–22 (2008)
2. Abdoli, M.A., Nazhad, M.F., Sede, R.S., Behboudian, S.: Longterm forecasting of solid waste
generation by artificial neural networks. Environ. Progress Sustain. Energy 00 (2011). doi:10.
1002/ep.10591
3. Batinic, B., Vukmirovic, S., Vujic, G., Stanisavljevic, N., Ubavin, D., Vumirovic, G.:
Using ANN model to determine future waste characteristics in order to achieve specific waste
management targets-case study of Serbia. J. Sci. Ind. Res. 70, 513–518 (2011)
4. Shahabi, H., Khezri, S., Ahmad, B.B., Zabihi, H.: Application of artificial neural network in
prediction of municipal solid waste generation (Case study: Saqqez City in Kurdistan
Province). World Appl. Sci. J. 20(2), 336–343 (2012). doi:10.5829/idosi.wasj.2012.20.02.
3769
5. Patel, V., Meka, S.: Forecasting of municipal solid waste generation for medium scale towns
located in state of Gujarat, India. Int. J. Innovative Res. Sci. 2(9), 4707–4716 (2013)
6. Roy, S., Rafizul, I.M., Didarul, M., Asma, U.H., Shohel, M.R., Hasibul, M.H.: Prediction of
municipal solid waste generation of Khulna city using artificial neural network: a case study.
Int. J. Eng. Res. Online 1(1), 13–18 (2013)
7. Falaahnezhad, M., Abdoli, M.: Investigating effect of preprocessing of data on the accuracy of
the modeling solid waste generation through ANNs. Amirkabir J. Sci. Res. (Civil and
Environmental Engineering) 46(2), 13–14 (2014)
8. Shamshiry, E., Mokhtar, M.B., Abdulai, A.M.: Comparision of artificial neural network
(ANN) and multiple regression analysis for predicting the amount of solid waste generation of
a tourist and tropical area—Langkawi Island. In: proceeding of International Conference on
Biological, Civil, Environmental Engineering (BCEE), pp. 161–166 (2014)
9. Singh, D., Satija, A.: Optimization models for solid waste management in indian cities: a
study. In: proceeding of fourth International Conference on Soft Computing for Problem
Solving, Advances in Intelligent Systems and computing (SocProS), vol. 335, pp. 361–371
(2014)
10. Pamnani, A., Meka, S.: Forecasting of municipal solid waste generation for small-scale towns
and surrounding villages located in state of Gujrat, India. International Journal of Current
Engineering and Technology 5(1), 283–287 (2015)
11. Rajashekaran, S., Pai, G.A.V.: Neural networks, fuzzy logic and genetic algorithms.
Rajashekaran, S., Pai, G.A.V., Ghosh A.K. (eds,) Fundamentals of Neural Networks,
pp. 11–30. Prentice Hall of India (2007)
292 Dipti Singh and Ajay Satija