Plane Thermoelastic Waves in Infinite Half-Space Caused Operational Research in Engineering Sciences: Theory and Applications Vol. 5, Issue 2, 2022, pp. 176-189 ISSN: 2620-1607 eISSN: 2620-1747 DOI: https://doi.org/10.31181/oresta190822135u * Corresponding author. anilutku@munzur.edu.tr (A. Utku), semakayapinar@munzur.edu.tr (Sema K.K) DEEP LEARNING BASED A COMPREHENSIVE ANALYSIS FOR WASTE PREDICTION Anıl Utku1, Sema Kayapinar Kaya2* 1 Department of Computer Engineering, Munzur University, Tunceli, Turkey 2* Department of Industrial Engineering, Munzur University, Tunceli, Turkey Received: 14 May 2022 Accepted: 27 July 2022 Type of paper Abstract: In its simplest definition, waste can be defined as any substance that is used, not needed and causes harm to the environment. Waste management covers control activities such as prevention of the formation of waste, reuse, separation according to its characteristics and type, storage, transportation, recycling and disposal. The main purpose of waste management is to leave a livable world to future generations, to create a sustainable environment, to protect natural resources, to save energy and costs, to reduce the rate of pollution and the amount of hazardous waste. In today's world where urbanization and industrialization rates are increasing, waste management is gaining importance. The aim of this study is to utilize waste data from Istanbul, Turkey's largest and fastest growing city, to estimate waste amount using a constructed Long Short-Term Memory (LSTM) based deep learning model. The developed LSTM-based model has been compared in practice with k-Nearest Neighbors (kNN), random forest (RF), Support Vector Machine (SVM), multi-layered perceptron (MLP) and Gated Recurrent Unit (GRU). As a result of the comparative and comprehensive analyzes, the experimental results showed that the developed LSTM- based deep learning method is more successful in the waste prediction problem than the other compared models. Key words: waste management, deep learning, machine learning, Long Short-Term Memory 1. Introduction Municipal solid waste (MSW) is presently one of the most pressing concerns in urban planning. MSW creation has accelerated as a consequence of global urbanization, population increase, and massive material consumption. Along with processes such as urbanization and urban transformation, the amount of municipal rubbish generated is increasing. The volume of MSW, one of the most significant by- products of an civil lifestyle, is increasing even faster than the growth of urbanization as the globe rushes towards an urban future. Today, MSW generation has grown to Deep Learning Based a Comprehensive Analysis for Waste Prediction 177 almost 3 billion residential per day, resulting in 1.2 kg of waste per person/day (1.3 billion tonnes). By 2025, this number is expected to rise to 4.3 billion urban communities, producing 1.42 kg/capita/day of municipal solid and around 2.2 billion tons per year (Hoornweg & Perinaz, 2012). Due to the large quantity of generated waste, an increasing amount of MSW can cause severe damage to the environment and population health (Huang et al. 2020). MSW management remains one of the best most tough problems for developing country governments to address to safeguard the environment and reduce public health risks. (Younes et al. 2016; Towa et al. 2020) and to maintain natural resources. To minimize the adverse effects of MSW, an accurate prediction of future waste volumes is a crucial issue. Efficient predicting of MSW generation is vital to the implementation of an optimum MSW management system as a primary tool for MSW management. Fu et al. (2015). Waste quantity predictions serve as the foundation for waste management process adoption, development, and optimization (Cubillos, 2020). Incorrect prediction of waste amounts can adversely affect process costs and cause systems to be inefficient. Another of the main challenges in using prediction models in waste management is the high variability and ambiguity of data wastage. The amount of waste produced can be affected by unpredictable factors such as people's behaviors, holidays, seasonal conditions. The volatile nature of waste data can cause problems in generating future forecasts. In addition, insufficient and incomplete data will reduce the accuracy of the predictions to be made. In recent years, significant studies on waste prediction using statistical modelling techniques, including machine learning (ML), have been presented (Abbasi et al. 2013; Abbasi & Hanandeh, 2016; Johnson et al. 2017; Ghanbari et al. 2021). The application of deep learning models to prediction issues has risen to prominence as high-performance data processing and computer power have advanced. Deep learning is a kind of ML that uses numerous layers of artificial neural networks to learn nonlinear interactions. It differs from typical ML approaches in that it allows for the learning of nonlinear relationships (Le et al. 2019). Due to the approaches' ability to understand common uncertainty and learn from long-term patterns, deep learning methods have proved to be a valuable technique for waste prediction and modelling. The Deep Learning (DL) has recently become popular in municipal waste generation (Sakr et al. 2016; Adedeji & Wang, 2019; Akanbi et al. 2020). In particular, LSTM models have proved to be successful in waste prediction problem, the efficiency of the forecasts is strongly dependent on the volume of historical waste dataset utilized to train the methods (Cubillos, 2020). In this study, a waste management model, which reduces cost and environmental damage, has been developed by using deep learning. Waste prediction enables efficient management of waste storage, logistics and disposal processes. It directly affects the ecosystem from factors such as waste management, climate change and air pollution. Therefore, it is very important to make accurate waste predictions. The growing quantity of solid waste generated by municipalities, as well as its disposal, has been one of Turkey's substantial environmental issues, particularly in Istanbul (Turan et al. 2009). Istanbul has a population of over 15 million people, or roughly 19% of Turkey's overall population. Istanbul generates an average of 18,000 tons of domestic waste every day. MW is collected regularly from different locations of the city by the Istanbul Environmental Management Industry and Trade Cooperation (ISTAC) company. Nearly 22,000 tons of MW have been compiled Utku et al./Oper. Res. Eng. Sci. Theor. Appl. 5 (2) (2022) 176-189 178 annually with the most advanced technological tools and specific outfits. In this study, it is aimed to predict the amount of waste produced by using the waste data in Istanbul, Turkey's largest and most developed city. The dataset used consists of the daily waste amounts in Istanbul, recorded for approximately 6 years between January 1, 2016 and October 31, 2021. The developed LSTM-based model has been compared in practice with kNN, RF, SVM, MLP and GRU. Experimental results obtained according to MSE (Mean Squared Error), RMSE (Root Mean Squared Error) and MAE (Mean Absolute Error) metrics for each model have been compared. To our knowledge, no other study has used a deep learning model for predicting Istanbul's MSW generation. The rest of the paper is presented as follows: The literature studies on waste prediction are covered in Section 2, which presents state-of-the-art methods for predicting and forecasting municipal waste. Data description and implementation of deep neural learning models are also presented in Section 3. Experimental results of the proposed LSTM model development and a comparison of metrics for several ML techniques are given in Section 4. Lastly, the discussion and conclusion are presented in sections 5. 2. Literature Review A diverse variety of prediction studies on waste generation rates have been studied by many researchers. Many different methods, such as traditional models, regression analysis, time series analysis, and artificial intelligence are some of the approaches utilized for waste prediction (Lavee & Khatib, 2010; Abbasi et al. 2013). The capacity to efficiently learn both linear and non-linear connections among time series and good prediction ability are two of those approaches' key benefits over traditional time series methods. Recently, ML algorithms have been employed successfully to predict waste generation (Xu et al. 2021). To predict waste generation using SVM combined with partial least square (Abbasi et al. 2013), a gradient boosting model (Johnson et al. 2017), two hybrid models based on decision trees and neural network was applied to predict Canada - wide municipal waste generation SVM and RF (Kumar et al. 2018), a hybrid model based on SVM and recurrent neural network (Meza et al. 2019), Gaussian process regression (GPR) model tuned by Bayesian optimization (Ceylan, 2020), prediction model using four combination intelligent algorithms, namely SVM, an integrated artificial neural network (ANN), RF, and multivariate adaptive regression splines (MARS) models (Ghanbari et al. 2021), decision tree and RF approach (Joshi et al., 2021). Among numerous nonlinear methods, ANN is one of the effective non-linear models to predict MSW generation. ANN could produce high-accuracy nonlinear estimation thanks to its intelligent learning method and hierarchical design. It has been successfully used in MSW prediction (Kannangara et al. 2018; Coşkuner et al. 2021; Abbasi et al. 2013; Jahandideh et al. 2009; Nguyen et al. 2021; Wu et al. 2020). Recently, the DL algorithms, a subset of ML algorithms, have recently shown huge success in a wide range of disciplines (Lecun et al. 2015; Kaya & Yıldırım, 2020). Deep learning allows models to construct hierarchical representations of incoming information with varying degrees of complexity. As a result, it can expose the complex structure of targeted information, enhancing pattern identification and Deep Learning Based a Comprehensive Analysis for Waste Prediction 179 classification capabilities (Lecun et al. 2015). One of the most important qualifications of deep learning architecture is that it can learn feature representations automatically, which saves time and effort. The rapid growth of deep learning can be attributed to improved chip processing capabilities, significant breakthroughs in ML algorithms, and the relatively low cost of computing hardware (Coşkun et al. 2017). Recently, DL approach has been widely used in waste generation at municipal level. Sakr et al. (2016) presented a combined convolution neural network (CNN) and SVM classification that categorized waste into 3 types: plastic, paper, and metal. They concluded that SVM had an accuracy of classification of 94.8 percent, while CNN only had an accuracy of 83 percent. SVM also demonstrated outstanding adaptability to various waste types. Adedeji and Wang (2019) suggested a new various waste classification system based on the 50-layer remaining net pre-train CNN model, which is a machine learning tool that is designed as an extractor, and SVM, which classifies waste into different groups/types such as glass, metal, paper, and plastic, among others. Akanbi et al. (2020) employed multi-layer deep learning architecture to create a computational tool for estimating the amount of construction materials and building demolition waste. Cubillos (2020) implemented a multi-site LSTMNN to anticipate garbage generation rates from residences. According to their results, LSTM model can successfully developed the results by 85% on average when compared to classical ML methods such as ARIMA. Wang et al. (2021) proposed a CNN to categorize waste into nine different garbage categories such as kitchen, other, plastic, glass, paper or cardboard, metal, fabric, and other recyclable waste. Furthermore, data exchange between garbage containers and the waste management point has been taken from the Internet of Things (IoT) sensors embedded in garbage containers. 3. Data Analysis Istanbul, which straddles the Bosporus strait and is in both Europe and Asia, has a population of over 15 million people, comprising nearly 19% of the total population, Istanbul is the most populated city in Europe and the fifteenth most crowded metropolis on the globe. Istanbul is the most populated city in Europe, and the world’s fifteenth-largest city. It is located on the Bosporus Strait, between the Black Sea and the Marmara Sea, in Turkey's northwest corner. Istanbul, Turkey is positioned at 41.015 latitude and 28.979 longitude. Istanbul, Turkey is in the urban place class of the Turkey country, with GPS coordinates of 41° 0' 54.493" N and 28° 58' 46.30" E. An average of 18,000 tons of household waste is generated daily in Istanbul. Approximately 5 thousand tons of household waste have been transported by district municipalities and 13 thousand tons by Istanbul Environmental Management Industry and Trade Cooperation (ISTAÇ) transport fleet to Solid Waste Transfer Stations located on Istanbul Asian side and European sides. ISTAÇ operates with a total of 8 thousand tons’ solid waste landfill on the European and Asian Sides of Istanbul, given in Fig. 1. Utku et al./Oper. Res. Eng. Sci. Theor. Appl. 5 (2) (2022) 176-189 180 Figure 1. Istanbul waste network’s structure Due to the population growth in Istanbul, the amount of waste is constantly increasing due to the increase in residential areas. Municipal waste collected from houses and workplaces is collected and separated by district municipalities. Transfer stations have been put into operation due to the cost and time consuming of direct transportation of the waste collected by the district municipalities to the landfills. In this way, fuel savings are achieved by eliminating the commuting of small-capacity municipal waste collection vehicles to landfills over long distances, traffic density is reduced; and possible air pollution caused by exhaust emissions is prevented. Then, the waste is brought to the sanitary landfills by transport trucks from the transfer stations. 3.1. Dataset In this study, an original statistical dataset consisting of daily waste amounts produced in Istanbul, recorded for approximately 6 years between Jan. 1, 2016 and Oct. 31, 2021, has been used. The dataset used consists of 2131 lines of waste data. The dataset contains the parameters of date and amount of waste produced. The first 10 rows of the dataset used are given in Figure 1. Fig. 2a shows the time distribution of the waste amounts in the dataset. Figure 2. Municipal waste amounts in the dataset Deep Learning Based a Comprehensive Analysis for Waste Prediction 181 The distribution of the amount of waste produced over time is given in Fig.2b. In addition, Fig. 2c shows the average amount of municipal solid waste produced between 2016 and 2021. 3.1. Developed Deep Learning Based Waste Prediction Model In this study, popular ML and deep learning models, that are widely used in the literature, such as kNN, RF, SVM, MLP, GRU and LSTM, are compared in practice. The dataset has been pre-processed before the models have been applied. Possible blank or incorrect fields in the dataset have been checked. After the data pre-processing step, training, validation, and test datasets have been selected. 80% of the dataset is split into training and 20% testing. 10% of the training data has been split for validation. Validation data has been used for the optimization of model parameters. For predictions to be made on time series data using machine learning and deep learning, it is necessary to structure the time series data as a supervised learning problem. For y to represent output and x to represent inputs, time series data needs to be converted to input-output prefixes y=f (x) using a function f. In supervised learning problems, it is expected to be possible to predict future data using past observation data. In supervised learning problems, it is to predict the value at time t by using the observation data at t-3, t-2 and t-1 time steps. In this study, using a 3- dimensional sliding window structure, the observation data at t-3, t-2 and t-1 time steps are configured as inputs and the observation value at time t as an output. In this way, time series data is transformed into a supervised learning problem. It is aimed to select the most suitable model parameters using validation data. Using the obtained parameters, the models have been created, and waste amounts have been predicted. The algorithm of the developed system is presented below: Figure 3. The algorithm of the developed system Utku et al./Oper. Res. Eng. Sci. Theor. Appl. 5 (2) (2022) 176-189 182 Deep learning models are artificial neural network models with complex architectures that aim to learn complex functions in high dimensions by making nonlinear transformations from input data to output data. LSTM is an advanced version of recurrent neural networks that stands out from other deep learning models. The main difference between RNN and LSTM is the retention time of information. LSTM is more advantageous than RNN because LSTM can process information in memory longer than RNN. LSTM cells retain cell states that have been read from and written to them. Based on the input and cell state values, the 4 gates regulate read, write and output to cell state. The developed LSTM based model is represented in Fig. 4. Figure 4. Developed LSTM-based deep learning model LSTM stands out over other deep learning models because of its success in remembering long-term dependencies. In this study, the developed LSTM-based deep learning model is compared with kNN, RF, SVR, MLP and GRU in practice. The developed deep learning model takes daily waste data as input and produces the predicted amount of waste as output. The developed LSTM based prediction model consists of an input layer, two LSTM layers, a dense layer and an output layer. In this study, it is aimed to optimize the parameters of the developed model. The time series data converted into a supervised learning problem structure is presented as an input to the LSTM. With parameter analysis studies, it is aimed to reach the highest prediction accuracy with parameters such as the number of layers, the number of neurons, the number of epochs and the batch size. Adam has been used as the optimizer. Deep Learning Based a Comprehensive Analysis for Waste Prediction 183 3.2. Experimental Results In this study, a dataset consisting of daily waste amounts produced in Istanbul, recorded for approximately 6 years between January 1, 2016 and October 31, 2021, has been used. The dataset contains the parameters of date and amount of waste produced. The developed LSTM-based deep learning model was compared against models such as kNN, RF, SVM, MLP, and GRU. MSE, RMSE and MAE values obtained for each model have been analyzed comparatively. The is represents the actual value of the waste amount at time t. The is the predicted waste amount at time t. The error value is calculated by . Scale-dependent metrics are effective when comparing different methods on the same dataset. Commonly used scale-dependent metrics are MAE and MSE metrics. The MAE metric is the mean of the errors as seen in Eq. (1). (1) MAE is the difference between the predicted values and the actual values. It is the mean of the absolute value of each difference between the actual value and the predicted value for that sample across all samples of the dataset. MSE metric has been calculated by mean of the squares of errors as seen in Eq. (2), and the RMSE has been calculated by the square root of the mean of the squares of error as seen in Eq. (3). (2) (3) Walk forward validation has been used in the applied models to eliminate the overfitting problem and improve the generated models' quality. All applied models have been run 10 times, and the average of the obtained results has been taken. The dataset used consists of 2131 lines of waste data. 80% of this data is split for training and 20% for testing. After the training/test split, 1704 lines of data have been used in the training and 427 rows of data have been used in the testing. Table 1 and Fig. 5 show the average MSE, RMSE and MAE results obtained for each model. Utku et al./Oper. Res. Eng. Sci. Theor. Appl. 5 (2) (2022) 176-189 184 Table 1. Experimental results for each model according to the MSE, RMSE and MAE Model MSE RMSE MAE kNN 1756420233222.19 1325300.05 978735.43 RF 1756605384968.68 1325369.90 978556.77 SVM 1722543245603.30 1312456.95 972004.19 MLP 1700317575744.18 1303962.26 932025.24 GRU 1672888061407.19 1293401.74 930672.17 LSTM 1670468166123.95 1292465.92 930325.64 The experimental results show that the average MSE values of kNN, RF, SVM, MLP, GRU and LSTM are 1756420233222.19, 1756605384968.68, 1722543245603.30, 1700317575744.18, 1672888061407.19, 1670468166123.95, respectively. The average RMSE values of kNN, RF, SVM, MLP, GRU and LSTM are 1325300.05, 1325369.90, 1312456.95, 1303962.26, 1293401.74, 1292465.92, respectively. The average MAE values of kNN, RF, SVM, MLP, GRU and LSTM are 978735.43, 978556.77, 972004.19, 932025.24, 930672.17, 930325.64, respectively. Figure 5. Comparative experimental results according to MSE, RMSE and MAE Experimental results showed that LSTM is more successful in waste prediction than kNN, RF, SVM, MLP and GRU. The LSTM's superior performance over other models is due to its architecture, which incorporates unique units in addition to the regular units found in the GRU. Fig. 6 shows the prediction results of LSTM on the test data. Deep Learning Based a Comprehensive Analysis for Waste Prediction 185 Figure 6. Prediction results of LSTM As seen in Fig. 6, the data marked in blue shows the test data, and the data marked in red shows the prediction values. The test dataset consists of 427 rows of data, that is, the amount of waste produced for 427 days. On the real data, the developed LSTM based prediction model shows a successful pattern. It is seen that SVM is more successful in waste prediction than kNN and RF. SVM requires less computation than kNN and is easier to interpret but can only describe a limited set of models. Also, kNN can find very complex patterns, but its output is more difficult to interpret. Both algorithms work better on categorical data. kNN is resistant to noisy training data and is effective in the case of a large number of training samples. RF works with a mixture of numerical and categorical features. RF is advantageous when features are of various scales. As a result, RF can utilize the data as is. SVM maximizes the distance between different points and calculates the distance between points. In the classification problem, RF gives the probability of belonging to the class, while SVM gives the points closest to the boundary between classes. Since the features in the data are numerical, SVM performed better than RF and kNN. SVM generally has higher prediction accuracy than MLP. SVM is usually better at prediction as there are advanced computations such as translating n-dimensional space using kernel functions. Neural network models require scaling of features. The numerical features in the dataset used in this study caused MLP to perform better than kNN, RF and SVM. As presented in Fig. 2c, the average amount of waste between 2016-2021 is 17894680.5. In Table 2, a comparative analysis of the MAE values, which express the average error values obtained from the applied models, according to the average waste amount values is presented. Utku et al./Oper. Res. Eng. Sci. Theor. Appl. 5 (2) (2022) 176-189 186 Table 2. Failure rates of applied models according to average waste amount and MAE values Model MAE Failure rates (%) kNN 978735.43 5.46 RF 978556.77 5.46 SVM 972004.19 5.43 MLP 932025.24 5.20 GRU 930672.17 5.20 LSTM 930325.64 5.19 Table 1 shows the failure rates according to the average waste amount and MAE values for the models applied. Equation 4 has been used to calculate the failure rate. (4) As seen in Table 1, the average waste amount and the failure value calculated according to the MAE values of the models have been calculated as 5.46 % for kNN, 5.46 % for RF, 5.43 % for SVM, 5.20 % for MLP, 5.20 % for GRU and 5.19 % for LSTM. 4. Conclusions The waste amount prediction problem is an important research area for waste management and recycling. Excessive population growth, combined with technological improvements and industrialization, are placing increasing pressure on the environment all around the world. While the development of production and marketing activities required a more intensive use of natural resources, the wastes created as a result of this trend have reached levels that endanger the environment and human health. The goal of this research is to establish an LSTM-based trash prediction model employing daily waste data. Using data from Istanbul's daily garbage, the constructed result was validated against kNN, RF, MLP, and GRU. The dataset used consists of the daily waste amounts produced in Istanbul, recorded for approximately 6 years between January 1, 2016 and October 31, 2021. The experimental results show that LSTM, GRU and MLP models have very successful results. Following these models, SVM, RF and kNN have been successful, respectively. The most successful results have been obtained with LSTM, and the most unsuccessful results have been obtained with kNN. The failure rate of LSTM is 5.19% while the failure rate of kNN is 5.46%. As a result of the comparative and comprehensive analyses, it has been seen that the LSTM-based deep learning model is applicable to the waste prediction problem. In the literature, there are studies in which machine learning and deep learning methods are used in waste management. However, there is no successfully implemented study to predict the amount of waste produced. In this study, the waste data of Istanbul, one of the largest industrial and tourist cities in the world, has been used for the first time in the literature. Experimental results showed that the deep Deep Learning Based a Comprehensive Analysis for Waste Prediction 187 learning-based prediction model can be successfully applied in industrial areas such as waste management. In future studies, more successful prediction results can be obtained by developing hybrid deep learning models. Acknowledgement: We thank the editor and anonymous reviewers for their constructive comments, which helped us to improve the manuscript. References Abbasi M, Abduli M.A., Omidvar B, Baghvand A. (2013). Forecasting Municipal Solid Waste Generation by Hybrid Support Vector Machine and Partial Least Square Model, 7, 27–38. https://doi.org/10.22059/IJER.2012.583. Abbasi M, El Hanandeh A. (2016). Forecasting municipal solid waste generation using artificial intelligence modelling approaches. Waste Management, 56,13–22. https://doi.org/10.1016/j.wasman.2016.05.018. Adedeji O, Wang Z. (2019). Intelligent waste classification system using deep learning convolutional neural network. Procedia Manuf., 35, 607–612. https://doi.org/10.1016/j.wasman.2016.05.018. Akanbi LA, Oyedele AO, Oyedele LO, Salami RO. (2020). Deep learning model for Demolition Waste Prediction in a circular economy. J Clean Prod., 274,.122843. https://doi.org/10.1016/j.jclepro.2020.122843. Ceylan Z. (2020). Estimation of municipal waste generation of Turkey using socio- economic indicators by Bayesian optimization tuned Gaussian process regression. Waste Management, 38, 840–850. https://doi.org/10.1177/0734242X20906877. Coşkun M, Yildirim Ö, Ayşegül U, Demir Y. (2017). An overview of popular deep learning methods. Eur J Tech EJT., 7, 165–176. Retrieved from https://dergipark.org.tr/en/pub/ejt/issue/34562/403498. Coskuner G, Jassim MS, Zontul M, Karateke S. (2021). Application of artificial intelligence neural network modeling to predict the generation of domestic, commercial and construction wastes. Waste Management Research, 39 (3), 499–507. https://doi.org/10.1177/0734242X20935181. Cubillos M. (2020). Multi-site household waste generation forecasting using a deep learning approach. Waste Management, 115, 8–14. https://doi.org/10.1016/j.wasman.2020.06.046. Daniel Hoornweg, Bhada-Tata-Perinaz. (2012). What a Waste: A Global Review of Solid Waste Management. Washington, District of Columbia: Urban Development & Local Government Unit World Bank. Fu H, Li Z, Wang R. (2015). Estimating municipal solid waste generation by different activities and various resident groups in five provinces of China. Waste Management, 41, 3–11. https://doi.org/10.1016/j.wasman.2015.03.029. Ghanbari F, Kamalan H, Sarraf A. (2021). An evolutionary machine learning approach for municipal solid waste generation estimation utilizing socioeconomic components. Arab J Geoscience, 14(2), 1-16. https://doi.org/10.1007/s12517-020-06348-w. Huang L, Cai T, Zhu Y, Zhu Y, Wang W, Sun K. (2020). LSTM-Based Forecasting for Urban Construction Waste Generation. Sustainability, 12 (20), 8555. https://doi.org/10.3390/su12208555. Jahandideh S, Jahandideh S, Asadabadi EB, Askarian M, Movahedi MM, Hosseini S, Jahandideh M. (2009). The use of artificial neural networks and multiple linear https://doi.org/10.22059/IJER.2012.583 https://doi.org/10.1016/j.wasman.2016.05.018 https://doi.org/10.1016/j.wasman.2016.05.018 https://doi.org/10.1016/j.jclepro.2020.122843 https://doi.org/10.1177%2F0734242X20906877 https://doi.org/10.1177%2F0734242X20935181 https://doi.org/10.1016/j.wasman.2020.06.046 https://doi.org/10.1016/j.wasman.2015.03.029 https://doi.org/10.3390/su12208555 Utku et al./Oper. Res. Eng. Sci. Theor. Appl. 5 (2) (2022) 176-189 188 regression to predict rate of medical waste generation. Waste Management, 29 (11), 2874–2879. https://doi.org/10.1016/j.wasman.2009.06.027. Johnson NE, Ianiuk O, Cazap D, Liu L, Starobin D, Dobler G, Ghandehari M (2017) Patterns of waste generation: A gradient boosting model for short-term waste prediction in New York City. Waste Management, 62, 3–11. https://doi.org/10.1016/j.wasman.2017.01.037. Joshi LM, Bharti RK, Singh R. (2021). Internet of things and machine learning-based approaches in the urban solid waste management: Trends, challenges, and future directions. Expert Systems, 39(5), 12865. https://doi.org/10.1111/exsy.12865. Kannangara M, Dua R, Ahmadi L, Bensebaa F. (2018). Modeling and prediction of regional municipal solid waste generation and diversion in Canada using machine learning approaches. Waste Management. 74, 3–15. https://doi.org/10.1016/j.wasman.2017.11.057. Kaya SK, Yıldırım Ö. (2020). A prediction model for automobiles sales in Turkey using Deep Neural Networks. J Ind Eng. 31, 57–74. Retrieved from https://dergipark.org.tr/en/pub/endustrimuhendisligi/issue/53786/669930. Kumar A, Samadder SR, Kumar N, Singh C. (2018). Estimation of the generation rate of different types of plastic wastes and possible revenue recovery from informal recycling. Waste Management, 79, 781–790. https://doi.org/10.1016/j.wasman.2018.08.045. Lavee D, Khatib M. (2010). Benchmarking in municipal solid waste recycling. Waste Management, 30, 2204–2208. https://doi.org/10.1016/j.wasman.2010.03.032. Le X-H, Ho HV, Lee G, Jung S. (2019). Application of Long Short-Term Memory (LSTM) Neural Network for Flood Forecasting. Water, 11 (7), 1387. https://doi.org/10.3390/w11071387. LeCun Y, Bengio Y, Hinton G. (2015). Deep learning. Nature, 521 (7553), 436–444. https://doi.org/10.1038/nature14539. Maroušek J. (2014). Economically oriented process optimization in waste management. Environment Science Pollution Research, 21 (12), 7400–7402. https://doi.org/10.1007/s11356-014-2688-z. Meza JKS, Yepes DO, Rodrigo-Ilarri J, Cassiraga E. (2019). Predictive analysis of urban waste generation for the city of Bogotá, Colombia, through the implementation of decision trees-based machine learning, support vector machines and artificial neural networks. Heliyon, 5(11), e02810. https://doi.org/10.1016/j.heliyon.2019.e02810. Nguyen XC, Nguyen TTH, La DD, Kumar G, Rene ER, Nguyen DD, Chang SW, Chung WJ, Nguyen XH, Nguyen VK. (2021). Development of machine learning - based models to forecast solid waste generation in residential areas: A case study from Vietnam. Resource Conserve Recycle, 167,105381. https://doi.org/10.1016/j.resconrec.2020.105381. Sakr GE, Mokbel M, Darwich A, Khneisser MN, Hadi A. (2016). Comparing deep learning and support vector machines for autonomous waste sorting. In 2016 IEEE international multidisciplinary conference on engineering technology (pp.127-142) .IEEE. 10.1109/IMCET.2016.7777453. Statista Demographies. In: Statista (2020). https://www.statista.com/statistics/1101883/largest-european-cities/. Accessed 21 April 2021. Towa E, Zeller V, Achten WM. (2020). Input-output models and waste management analysis: A critical review. Journal of Clean Production, 249, 119359. https://doi.org/10.1016/j.jclepro.2019.119359. https://doi.org/10.1016/j.wasman.2009.06.027 https://doi.org/10.1016/j.wasman.2017.01.037 https://doi.org/10.1111/exsy.12865 https://doi.org/10.1016/j.wasman.2017.11.057 https://doi.org/10.1016/j.wasman.2018.08.045 https://doi.org/10.1016/j.wasman.2010.03.032 https://doi.org/10.3390/w11071387 https://doi.org/10.1007/s11356-014-2688-z https://doi.org/10.1016/j.heliyon.2019.e02810 https://doi.org/10.1016/j.resconrec.2020.105381 https://doi.org/10.1109/IMCET.2016.7777453 https://doi.org/10.1016/j.jclepro.2019.119359 Deep Learning Based a Comprehensive Analysis for Waste Prediction 189 Turan NG, Çoruh S, Akdemir A, Ergun ON. (2009). Municipal solid waste management strategies in Turkey. Waste Management, 29(1), 465-469. https://doi.org/10.1016/j.wasman.2008.06.004. Wang C, Qin J, Qu C, Ran X, Liu C, Chen B. (2021). A smart municipal waste management system based on deep-learning and Internet of Things. Waste Management, 135, 20–29. https://doi.org/10.1016/j.wasman.2021.08.028. TUIK. (2021). https://data.tuik.gov.tr/Bulten/Index?p=Adrese-Dayal%C4%B1- N%C3%BCfus-Kay%C4%B1t-Sistemi-Sonu%C3%A7lar%C4%B1-2020- 37210&dil=1. Accessed 21 April 2021. Waste Management in Istanbul. (2015). https://atikyonetimi.ibb.istanbul/hizmetlerimiz/kati-atik-aktarim-istasyonlari/. Accessed 30 December 2021. Where is Istanbul, Turkey on Map Lat Long Coordinates (2021). https://www.latlong.net/place/istanbul-turkey-2242.html. Accessed 21 December 2021. Wu F, Niu D, Dai S, Wu B. (2020). New insights into regional differences of the predictions of municipal solid waste generation rates using artificial neural networks. Waste Management, 107, 182–190. https://doi.org/10.1016/j.wasman.2020.04.015. Xu A, Chang H, Xu Y, Li R, Li X, Zhao Y. (2021). Applying artificial neural networks (ANNs) to solve solid waste-related issues: A critical review. Waste Management, 124, 385–402. https://doi.org/10.1016/j.wasman.2021.02.029 . Younes MK, Nopiah ZM, Basri NEA, Basri H, Abushammala MFM, Younes MY. (2016). Landfill area estimation based on integrated waste disposal options and solid waste forecasting using modified ANFIS model. Waste Management, 55, 3–11. https://doi.org/10.1016/j.wasman.2015.10.020. © 2022 by the authors. Submitted for possible open access publication under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). https://doi.org/10.1016/j.wasman.2008.06.004 https://doi.org/10.1016/j.wasman.2021.08.028 https://doi.org/10.1016/j.wasman.2020.04.015 https://doi.org/10.1016/j.wasman.2021.02.029 https://doi.org/10.1016/j.wasman.2015.10.020