جستجوی مقالات مرتبط با کلیدواژه "ماشین بردار پشتیبان" در نشریات گروه "محیط زیست"

The most important goal in the planning and optimal operation of the reservoir system is to determine the various operating policies that can operate properly Under the drought condition and existing uncertainties. In this study, a combination of the Horse herd optimization algorithm (HOA) and WEAP simulator model was used to extract the optimal reservoir exploitation policies in the form of specific optimization and objective functions were calculated based on the results of the implementation of each scenario and the total operating period for Maroon and Jarreh reservoir dams. The results showed that the average error of the optimal rules extracted from the support vector machines relative to the output of the HOA algorithm in the validation stage is less than 17%, which indicates the high efficiency of this method in predicting the optimal pattern of the dam control curve in real-time. Moreover, evaluation of different scenarios showed that agricultural development in areas 1, 4, and 5 of Ramhormoz will be reduced by an average of 50% and also a 10% reduction in inflow to Marun and Jarreh dams will have negative effects on Shadegan wetland.

In this study, geostatistical methods and artificial intelligence models (artificial neural network, decision tree, and support vector machine) were used to simulate the soil salinity of Ghorghori lands in Hirmand city. A total of 130 soil samples were collected from 0-30 cm layers of the soil. The electrical conductivity of each sample was measured using an electrical conductivity device. Soil salinity values were estimated using Geostatistical methods and artificial intelligence methods. Geostatistical and artificial intelligence models were applied and the best model was selected; the accuracy of the methods was compared using independent validation. The results showed that the artificial intelligence methods outperformed the geostatistical method in estimating the soil salinity of the artificial intelligence methods, the decision tree model was the superior model due to its coefficient of determination of 0.99 and RMSE and MAE statistics of 0.26 and 0.18 respectively. The salinity trend showed that the salinity of the soil of the region decreases from west to east first and then increases and decreases from north to south. In order to preserve the environment of the region, the field of planting plant species compatible with the region should be provided in accordance with soil salinity.

Urban growth has accelerated in recent decades, therefore, predicting the future growth pattern of the city is very important to prevent some environmental, economic and social problems. The city of Tabriz has also experienced rapid growth of urban lands due to significant demographic changes, which requires accurate simulation of urban growth to prevent negative environmental and economic consequences. The purpose of this study is to evaluate the performance accuracy of the proposed machine learning algorithms by spatial cross-validation method in combination with the cellular automata model to simulate urban growth.

In this study, to analyze urban land use changes, Landsat satellite images related to the years 1997, 2006 and 2015 were classified using the support vector machine algorithm. In the next step, change potential maps of non-urban to urban areas using random forest algorithms, support vector machine and multilayer perceptron neural network for two periods of calibration (1997 and 2006) and validation (2006 and 2015) based on distance from the main roads, distance from the city center, distance from built-up areas, distance from the rivers and railways, as well as slope, elevation and two-class (agricultural / barren) land use layer were produced as effective factors in the growth of the city. Finally, using the cellular automata model, the growth simulation of Tabriz city based on land use and change potential maps obtained from machine learning algorithms for the mentioned periods was performed. To prevent over-fitting of algorithms to training samples and to obtain optimistic results, in the process of extracting optimal parameters of machine learning algorithms, the spatial cross-validation method was used to reduce spatial correlation between training and test data.

The results showed that the random forest algorithm with the area under the ROC curve of 0.9228 compared to the support vector machine and multilayer perceptron neural network algorithms with 0.8951 and 0.8726, respectively, had a better performance in estimating the change potential of non-urban to urban areas. Furthermore, in comparison with others, random forest also clearly showed local variations in potential change. Finally, the growth of Tabriz city was simulated using the cellular automata model based on the obtained change potential maps. Comparison of the prediction map in the validation period with the current situation of urban areas in 2015 showed that the accuracy of urban growth simulation model based on random forest with a Figure of Merit index of 0.3569 compared to models based on support vector machine and artificial neural network was more accurate in allocating non-urban to urban lands with 0.3496 and 0.3434, respectively.

As machine learning algorithms such as artificial neural network, support vector machines and random forest are capable of solving non-linear problems, using them is strongly recommended for urban growth simulation. Also, among the algorithms used in this research, the random forest algorithm based on ensemble learning has a higher advantage than the two support vector machine and the artificial neural network algorithms.

National parks and wildlife shelter are the most important natural heritages; therefore, knowing of quantitative and qualitative changes in their land use plays an essential role in the quality of these areas' management. various algorithms have been developed to classify satellite imagery in remote sensing, selecting an appropriate classification algorithm is very important in achieving the accurate results. In this research, a more accurate algorithm was determined by comparing the classification accuracy of two artificial neural network and support vector machine algorithms, and it was used to examine the process of the land use changes. The present study was performed in Boujagh National Park, in the Guilan Province, during the years 2000 to 2017, using satellite imagery ETM and OLI of Landsat 7 and 8. The results of the research revealed that the support vector machine algorithm with overall accuracy and Kappa coefficient of 86.42 and 0.83 respectively for the year 2000 and, 90.65 and 0.88 for the year 2017, classified the satellite images more precisely, in comparison with the artificial neural network algorithm with overall accuracy and Kappa coefficient of 83.71 and 0.80 respectively for the year 2000 and overall accuracy and Kappa coefficient of 89.25 and 0.87 for the year 2017. Therefore, the land use maps of the support vector machine algorithm were used to determine the land use changes. The study of land use change by this method concluded that the areas of the waterbody, sea, grassland and agriculture have decreased and marshland, woody and bare lands classes showed an increase during the study period.

The aim of this study was to determine the accuracy of Landsat 8 and Sentinel 2 satellite data sets based on Maximum Likelihood, Support Vector Machine, Neural Network algorithms for mapping the LU/LC of Kobgian watershed in Kohgiluyeh and Boyer-Ahmad Province. For this purpose, corrections, data preparation, data set creation, classification and analysis, mapping and verification were done using ENVI® 5.3, ArcGIS® 10.5, Google Earth Pro and Excel 2016 software. The results showed that the highest total accuracy and kappa coefficient for Landsat 8 and Sentinel 2 satellites belongs to the maximum likelihood algorithm with a value of 74.18% and 0.69 and neural network algorithm with a value of 72.84% and 0.67, respectively. The overall accuracy order of the algorithms for mapping LU/LC the watershed using Landsat 8 and sentinel 2 data was as maximum likelihood > support vector machine > neural network and using data was as neural network> maximum likelihood > support vector machine, respectively. The accuracy of the algorithms indicated that if a specific LU/LC is the main goal such as basin rangelands, the support vector machine algorithm should be used. The area of eight classes of Kabgian watershed is: agriculture 1342, residential 1356, rock 3579, forest 23289, water body 407, abandoned lands 9571, garden 3139 and pasture 54125 ha. Therefore, depending on the type of LU/LC, the type of satellite data available, and the purpose of study, the priority of using algorithms will be different and based on the desired factor, suitable algorithm should be selected.

In many studies concerning monitoring the water surface changes, multi-temporal images are sepa-rately analyzed and after extracting water boundaries in each image, these boundaries are compared and the changes are determined. Nevertheless, there is a demand for methods that can perform accurately as well as facilitating the identification of changes. Therefore, to this end, in this research synergy of multi-temporal image fusion and classifi-cation methods were investigated to detect surface water changes in Maharlu Lake between 2013 and 2018.

After performing the necessary pre-processing, the Gram-Schmidt (GS) and Principal Component Analysis )PCA( methods were applied to fuse images and, then, changed and unchanged areas were extracted through applying classification methods to the fused images. Support Vector Machine )SVM) and Max-imum Likelihood )ML( were used to classify fused images. In the next step, combinations of these methods were compared to each other and the best pair was extracted. Finally, the selected pair was compared with conventional change detection methods.

The results showed that based on the GS-SVM methods, the Maharlu Lake retreated about 163.3 km2 from 2013 to 2018. For accuracy assessment of the methods, the overall accuracy and Kappa coefficient were calculated. The GS-SVM method had an overall accuracy of 99.33%, Kappa coefficient of 0.99 and a relative error of 3.92 km2. This pair detected changes more accurately and the results were closer to reality. In the next step, the water surface was extracted from the images using conventional change detection methods, such as image differencing, band rationing, and NDVI differencing, and their results were compared to that of the GS-SVM. According to the results, the GS-SVM had higher overall accuracy and Kappa coefficient compared with other methods, and simultaneously, the least relative error.

The results of this study showed that a combination of the GS image fusion method and SVM clas-sifier provides satisfactory results to extract changes from multi-temporal images. This synergy can be used as an effective tool for detecting changes, particularly since fusing images can also be effective in improving classifica-tion accuracy by enhancing the spatial resolution of images.

The first step in design of municipal waste management systems is complete understanding of waste generation quantity. Forecasting waste generation is one of the most complex engineering problems due to the effect of various and out of control parameters on waste generation. Therefore, it is obvious that it is necessary to develop approaches to a model such complex events. The objective of this study is forecasting waste generation quantity using intelligent models as well as their comparisons and uncertainty analysis.

In this study, Mashhad city was selected as a case study and waste generation time series of waste generation in 1380 to 1390 were used for weekly prediction. Intelligent models including artificial neural network, support vector machine, adaptive neuro-fuzzy inference system as well as K-nearest neighbors were used for modelling. After optimizing the models’ parameters, models’ accuracy were compared by statistical indices. Finally, result uncertainty of the models was done by Mont Carlo technique.

Results showed that coefficient of determination (R2) of artificial neural network adaptive neuro-fuzzy inference system, support vector machine, and K-nearest neighbor models were 0.67, 0.69, 0.72 and 0.64 respectively. Uncertainty analysis was also justified the results and demonstrates that support vector machine model had the lowest uncertainty among other models and the lowest sensitivity to input variables.

Intelligent models were successfully able to forecast waste quantity and among the studied models, support vector machine was the best predictive model. Moreover, support vector machine produced the results with the lowest uncertainty the other models.

Having information about qualitative and quantitative parameters distribution of groundwater supplies is one of most important parameters in integrated groundwater management. Thus, in this study it has been attempted to determine a proper model and input combination for estimation of quality parameters including electrical conductivity (EC), calcium (Ca) and sodium (Na) ions in aquifers of Guilans plain. In this study, the data from 132 observation wells during 2001 to 2013 were used and artificial neural network (ANN) and support vector model (SVM) were applied. In the first approach, estimations were conducted according to five different combinations, including water level, distance from see, total precipitation of six months and coordinates of observation wells. In the second approach, estimations were conducted based on combination of the selected qualitative parameters of gamma test with combinations of the best input in the first part. Comparison of the results from the first part indicated that SVM model outperformed the ANN mode in the estimation of Ca, Na and EC parameters. Support vector machine error values for estimating Ca, Na and EC variables at the test period were 1.218 (meq/l), 0.867(meq/l), and 175.742 (µmos/cm), while for artificial neural network these values were 1.268 (meq/l), 0.933 (meq/l), and 186/448 (µmos/cm) respectively. The results from this part showed that adding the distance from see input improves the estimation of models in all cases. In the second part, using gamma test for measuring the nine quality parameters, the best combination of quality parameters was determined to estimate the three parameters: Ca, Na and EC. The results from the second part show that both ANN and SVM models have an excellent performance in the estimation of the three qualitative parameters. ANN model error values in estimating Ca, Na and EC variables in validation period were 0.662 (meq/l), 0.305(meq/l), and 47.346 (µmos/cm), while these values were 0.671 (meq/l), 0.356 (meq/l), and 55.412 (µmos/cm) for SVM model respectively.  Obviously, the results from ANN model in this section were better than those from SVM model. Results showed that both ANN and SVM models have a great ability in predicting qualitative parameters in the aquifers. Also, in less inputs, the results of SVM model are better than those of ANN model and in more inputs it is vice versa. Results of the second section showed that gamma test is fully practical and accurate in determining the effective input combinations.

Prediction of river flow is an important issue in planning water resources and management of supply and demand in future conditions. Hence, it has attracted researchers’ notice due to its importancein the designing, planning, management and operation of water facilitiesand also management the critical conditions such as flood and drought. In the present study, it was attempted to improve Zarinehrud river inflow prediction for use in water resource planning using a hybrid approach based on gamma test and supporting vector machine model (GSVM). For this purpose, the best possible combination of predictors was selected from the different combinations of 10 meteorological and hydrological variables in the basin. Then, based on the best combination of predictors, the potential of river inflow was predicted using a support vector machine. Comparison of predicted and observed flow indicated the good performance of hybrid approach in prediction of potential river inflow for application in basin management plans. In this case, the overall accuracy of the model to predict drought management levels based on Zarinehrud flow is 71.4%, and the upper and under estimation error are 8.2 and 20.4% respectively. These results show the acceptable precision of GSVM model for flow prediction in different hydrological situations of basin.
Application of Hybrid Gamma-SVM Approach for River Flow Prediction in Zarinehrud Basin

IntroductionFive-day biochemical oxygen demand (BOD5) is one of the key parameters which is widely applied to evaluate the biological and chemical conditions of rivers, and to study the effects of releases in rivers from wastewater treatment plants, factories, farms, etc. Measurement of this parameter requires about 5 days. However, 5 days is a long time for the determination of BOD5. Therefore, in the present research two software sensors, i.e. artificial neural network (ANN) and support vector machine (SVM) are developed for the online estimation of BOD5. To achieve this goal, the Sefidrood River in Iran was selected. It is pointed out that although ANN models have been used to predict BOD5, but using SVM techniques for the online estimation of this parameter have never been reported.Materials and Methods ANN and SVM models are two well known artificial intelligent techniques which have frequently been used for modeling nonlinear parameters in engineering problems. In this research, a feed-forward ANN technique and a regression model is considered for online prediction of BOD5. To optimize ANN parameters, three optimization algorithms, i.e. Levenberg-Marquardt (LM), resilient back-propagation (RP), and scaled conjugate gradient (SCG) are used. Besides, two-steps grid search algorithm is considered for the optimization of SVM parameters, i.e. ε, C and γ.Results and DiscussionTo achieve the best network geometry according to the highest correlation (R), and the lowest root mean square error (RMSE) of the testing sets for online estimation of BOD5, various structures of NN model with different training functions are investigated. Finally, the network with LM algorithm (ANN (LM) model) is selected as the best ANN model. The results of training and testing steps for developed ANN models with different training functions are given in Table 1.In addition, for SVM model development, first, ε, C and γ should be determined. In this research, the best fitting value was obtained by trial and error procedure. Therefore, this parameter was set to several values. Thus, 14 models were developed (ε and C are considered to be varied). The RMSE value has been assessed on different values in the different -SVM regression models. Obtained results indicated that the SVM model including the value equal to 1.472 has the least RMSE. Therefore, it can be chosen as the best SVM model for the online prediction of BOD5. However, optimal values for C and using two-step grid search were 13 and 0.037, respectively. Finally SVM model is trained and tested using the determined best fitted values.