meta heuristic algorithms

This research aimed to select essential features for modeling the cost of pressurized irrigation systems using the data of 515 drip irrigation projects in four parts, including the cost of pumping station and central control system (TCP), cost of on-farm equipment (TCF), cost of installation and operation on-farm and pumping station (TCI), and total cost (TCT). In the first stage, a database including 39 features influencing the cost of the mentioned sectors was prepared and the price of all projects (2006 to 2019) was updated for the base year of 2021. Then, feature selection was done with different algorithms in MATLAB environment and in two parts including (1) all features (39 features before and after the design stage) and (2) 18 features before the design phase (BD). The results showed that the amounts of RMSE and R2 for all the features were equal to 0.007 and 0.92, respectively, and for the BD section, they were equal to 0.003 and 0.89, respectively. Among the different algorithms for feature selection, support vector machine (SVM) and optimization algorithms (Wrapper) were identified as the best learner and feature selection method, respectively. The results of the evaluation criteria showed that the two LCA and FOA algorithms achieved the best estimation, and their error criterion in all the features were 0.0020 and 0.0018, respectively, while their correlations were 0.94 and 0.94. In the BD features, these criteria were 0.0006 and 0.95 for both algorithms, respectively. Finally, in the all features section, 10 out of 39 features and for BD section, 8 out of 18 were selected as the most effective features. The results of choosing the most effective features that affect the cost of different parts of the drip irrigation system can make the cost modeling of the systems simpler and faster and, while being useful for research works, it facilitates estimation and management of costs before implementation of each project.

Accurate prediction of pore water pressure in the body of earth dams during construction with accurate methods is one of the most important components in managing the stability of earth dams. The main objective of this research is to develop hybrid models based on fuzzy neural inference systems and meta-heuristic optimization algorithms. In this regard, the fuzzy neural inference system and optimizing meta-heuristic algorithms including genetic algorithms (GA), particle swarm optimization algorithm (PSO), differential evolution algorithm (DE), ant colony optimization algorithm (ACOR), harmony search algorithm (HS), imperialist competitive algorithm (ICA), firefly algorithm (FA), and grey wolf optimizer algorithm (GWO) were used to improve training system. Three features including fill level, dam construction time, and reservoir level (dewatering) obtained from the dam instrumentation were selected as the inputs of hybrid models. The results showed that the hybrid model of the genetic algorithm in the test period had the best performance compared to other optimization algorithms with values of R2, RMSE, NRMSE, and MAE equal to 0.9540, 0.0866, 0.1232, and 0.0345, respectively. Also, ANFIS-GA, ANFIS-PSO, ANFIS-ICA, and ANFIS-HS hybrid algorithms performed better than ANFIS-GWO, ANFIS-FA, ANFIS-ACORE, and ANFIS-DE in improving ANFIS network training and predicting pore water pressure in the body earthen dams at the time of construction.

Introduction :

Sugarcane is a tropical, perennial grass that forms lateral shoots at the base to produce multiple stems. It is the main source of sugar production and one of the most important sources of energy production in the world. Today, the use of artificial intelligence and data mining findings to help predict product production is considered. Determining the relationship between inputs and outputs of production process using artificial intelligence (AI) has drawn more attention rather than mathematical models to find the relationships between input and output variables by training, and producing results without any prior assumptions. The adaptive neuro-fuzzy inference system (ANFIS), as a form of AI, is a combination of artificial neural network (ANN) and fuzzy systems that uses the learning capability of the ANN to derive the fuzzy if-then rules with appropriate membership functions worked out from the training pairs, which in turn leads to the inference.Particle swarm optimization (PSO) is an algorithm modeled on swarm intelligence, in a search space, or model it finds a solution to an optimization problem and predict social behavior in the presence of objectives. The PSO is a population-based stochastic computer algorithm, modeled on swarm intelligence. Swarm intelligence is based on social psychological principles and it provides insights into social behavior, also helps to many engineering applications. Feature selection is becoming very important in predictive analytics. Indeed, many data sets contain a large number of features, so we have to select the most useful ones. One of the most advanced methods to do that is the genetic algorithm (GA). Genetic algorithms can select the best subset of variables for predictive model. The purpose of this research is to evaluate the applicability of one artificial intelligence technique including adaptive neuro-fuzzy inference system and also combining this technique with particle swarm optimization to increase the accuracy and speed of training of the neuro-fuzzy system in prediction of yield and recoverable sugar percentage (R.S%) of sugarcane.

Materials and Methods:

 In this paper, one main pattern of adaptive neuro-fuzzy inference system (ANFIS) and one synthetic model of adaptive neuro-fuzzy inference system with particle swarm optimization (PSO) were used to predict the studied properties by MATLAB version 2017. Initial data for this study were collected from Debal-Khozaie Agro-industry Company in Khouzestan province, Iran. The actual data for the seven periods of sugarcane harvest from 2010 to 2017 were used for modeling. The studied parameters included a set of agronomic factors, soil characteristics, irrigation and climate in the study area. The test data sets were used for comparison of selected ANFIS and ANFIS with PSO, as well as for the observation values. This comparison was performed by using three statistical indices: Determination Coefficient (R2), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE).

Results and Discussion:

From all of the studied parameters, eleven parameters were selected as the effective features by the binary genetic algorithm (BGA). In feature selection, the function to optimize is the generalization performance of a predictive model. More specifically, in this method, purpose was to minimize the error of the model on an independent data set not used to create the model. The data were randomly divided into two groups: training and testing. Each pattern was modeled separately and then the results were compared. The results showed that the combination of adaptive neuro-fuzzy inference system with particle swarm optimization algorithm (ANFIS-PSO) had better performance in predicting cane yield and recoverable sugar percentage. In ANFIS-PSO model the root mean square error, mean absolute percentage error and coefficient of determination values were found 0.0181, 0.0217, 0.9237 and 0.0086, 0.0138, 0.9847 respectively for two variables of cane yield and recoverable sugar percentage. In relation to the predicted cane yield by the neuro-fuzzy network with particle swarm algorithm, it can be concluded that among the effective factors, with increasing plant age and use of resistant varieties, the amount of yield was decreased and increased, respectively.

Conclusion:

 The hybrid pattern of adaptive neuro-fuzzy inference system with the particle swarm optimization has been directed against the mere neuro-fuzzy system to a more accurate and stronger solution. Indeed, it can be concluded that ANFIS model with the PSO has the ability for precise estimation of sugarcane yield and recoverable sugar percentage.

Due to the flow regime and consequently the sediment regime are not constantly in the watersheds, the prediction of sediment discharge is a great help in estimating and managing the sediment input to hydraulic structures. Measurement of sediment in the usual way is not justified in nowadays and may also lead to human error. Therefore, in this study, three meta-heuristic optimization algorithms, including imperialist competitive algorithm (ICA), grey wolf optimizer algorithm (GWO) and election algorithm (EA), were used to predict the suspended sediment load of the Zarrineh river. In order to calculate the sediment discharge by the models, firstly, the necessary statistics and data were collected from the studied station in the period 1993-2015. After processing the data, 210 corresponding discharge and sediment data were selected. The corresponding discharge-sediment data from the study station were randomly separated into two parts, 70% for training and 30% for testing. In order to evaluate the performance of the algorithms, four statistics consist of R2, RMSE, MAE and the NSE were used. The results showed that GWO algorithm with values of statistical criteria R2=0.96, RMSE=228.86 ton/day, NSE=0.74 and MAE=67.32 ton/day, has a very high accuracy compared to other algorithms used which this would lead to comprehensive planning for the design and construction of hydraulic structures.

In this study, to predict sugarcane yield, an Artificial Neural Network based on meta-heuristic algorithms is used as an efficient method to estimate crop yield based on actual data. In order to predict sugarcane yield, effective parameters such as plant characteristics, electrical conductivity of soil and water, maximum temperature and average hours of sunshine during the growing season and on a time scale of seven years were used. Accordingly, four hybrid models were used to build neural networks which including of artificial neural network based on Back Propagation (BP) algorithm, combining neural network with Genetic Algorithm (GA), combining neural network with Particle Swarm Optimization (PSO) algorithm and finally a new approach as combining neural network with GA-PSO. The results show that the neural network performance can be improved using meta-heuristic algorithms and can be increased significantly the prediction power. The Mean Square Error (MSE) and correlation coefficient (R) in the hybrid method of neural network with GA-PSO were obtained 0.00057 and 0.91457 respectively on the test data, which show the superiority of this method to other patterns. In addition, cross validation test of the proposed model by K-Nearest Neighbors showed the accuracy of training and test data to predict sugarcane yield has been equal to 98.5% and 95.5%, respectively.

In many areas of Iran, there is no detailed information on the amount of erosion, sediment transport and sedimentation of rivers, and in many cases, there are many difference between measurements. Due to the fact that the flow regime and consequently the sediment regime in the watersheds are not constant, the prediction of sediment rate helps to estimate the sediment accumulated behind the structures, specially the dams, and determine the dead volume of reservoirs in the future months, and by adopting timely arrangements facilitate the deposition management to a certain extent. In this research, three optimization algorithms including Genetic Algorithm (GA), Gray Wolf Optimizer (GWO) and Election Algorithm (EA) were used to predict the suspended sediment load of the rivers. In order to evaluate the performance of the algorithms, three statistics consists of R2, RMSE and NSE were used. The suspended sediment load of sedimentary station located in the Zarrineh-Rood river during the 2005-2015 are used as a case study. The results show GWO algorithm with values R2=0.96, RMSE=0.022 and NSE=0.74 has a very high accuracy compared to other algorithms used.

Use of the curve gradient of the Soil Water Retention Curves (SWRC) in the inflection point (S Index) is one of the main indices for assessing the soil quality for management objectives in agricultural and garden lands. In this study Anneling Simulated – artificial neural network (SA-ANN) hybrid algorithm was used to identify the most effective soil features on estimation of S Index in Jiroft plain. For this purpose, 350 disturbed and undisturbed soils samples were collected from the agricultural and garden lands and then some physical and chemical soil properties including Sand, Silt, Clay percent, Electrical Conductivity at saturation, Bulk Density, total porosity, Organic Mater, and percent of equal Calcium Carbonate were measured. Moreover, the soil moisture amount was determined within the suctions of 0, 10, 30, 50, 100, 300, 500, 1000, 1500 KP using pressure plate. Then, the determinant features influencing the modeling of S Index were derived using SA-ANN hybrid algorithm. The results indicated that modeling precision increased by reducing the input variables. According to the sensitivity analysis, the Bulk Density had the highest sensitivity coefficient (sensitivity coefficient=0.5) and was identified as the determinant feature for modeling the S Index. So, since increasing the number of features does not necessarily increase the accuracy of modeling, reducing input features is due to cost reduction and time-consuming research.

This study was aimed to investigate the best features extracted from images to determine the best technique for qualitative classification of green tea by using meta-heuristic algorithms. Five different classes of green tea were evaluated according to the standards of the Institute of Standards and Industrial Research of Iran. After receiving the images of different green tea classes in the computer, 6 square blocks were extracted from each of the original color images. These image blocks were transformed from RGB to gray scale images. One-level discrete Haar wavelet filter was applied to the gray images and 4 wavelet subimages were obtained. Co-occurrence matrices were calculated for each wavelet subimages and 17 common texture features in the image textural studies, were extracted from subimages (totally 68 texture features for each block image). By using principal component analysis, 8 feature components were produced from the original features and used for the separation of 5 groups of green tea. The results showed that algorithms of artificial neural networks, support vector machine and decision tree were capable of qualitative classification of green tea with high accuracy. However, Bayesian network did not have acceptable performance. According to the evaluation statistics, the multilayer perceptron artificial neural networks (with Kappa statistic, root mean square error and classification accuracy of 0.90, 0.42, and 99.17%, respectively) was the best classifier. Based on the results of this study, the use of machine vision and texture features extracted from image wavelet subimages is a suitable technique for the qualitative classification of green tea.

The weeds must be removed from the field due to their competition with principal crops to use water, nutrients, sunlight, etc. There are different methods to remove the weeds: mechanically, manually or chemically (applying herbicides). For farmers, applying herbicides is a usual way, but brings some concerns, from the point of environmental issues, due to equal application of chemicals all over fields, regardless the presence or absence of weed. For this reason, a machine vision system based on video processing was proposed to recognize Secale cereale L. (as a weed) from potato plant (as principal crop) to make herbicide application more accurate. Nine hundred sixty five objects were recognized after taking videos, pre-processing and segmentation. Fourteen features were extracted from each object. Using the hybrid artificial neural network-genetic algorithm, of 14 extracting features, only 6 features were selected as effective ones: average, the third moment, autocorrelation, correlation, dissimilarity, and entropy. Data were classified into two groups: training data (70% of the total data) and testing data (30% of the total data). The classification was performed using hybrid of artificial neural network - Bio-geography Based Optimization (BBO) algorithm. Performance of classification system was evaluated through analysis of confusion matrix and Receiver Operating Characteristic (ROC). Sensitivity, specificity, and accuracy were calculated using confusion matrix. The results showed that the sensitivity, accuracy and specificity of classification system reached to an acceptable level: 99.49 %, 99.65% and 98.91%, respectively. Our conclusion is that it is possible to manufacture the machine vision system with mentioned aims that work as online.

Long-term groundwater monitoring networks can provide essential information for planning and management water resources. Budget constraints in water resource management agencies often mean a reduction in the number of observation wells included in a monitoring network. Due to economic considerations and reducing monitoring costs, the optimization approach in this study is to reduce the monitoring stations. In the present study, genetic algorithm was used to designing optimal water level monitoring network. The aim of optimization to determine an optimal combination (reduced) from the main network was well observed, so that the possibility of measurement error smallest and least loss of data and provide the best distribution of wells. Using the genetic algorithm, the groundwater monitoring network of the region was optimized. Using the genetic algorithm, the groundwater monitoring network of the region was optimized. In Koohdasht Plain, among of 15 observation wells, 13 wells observations had data. in order to do this research, the data for 36 consecutive months, namely, data from 1392 to 1394 were used. Using the genetic algorithm, the groundwater monitoring network was redesigned and selected wells were selected from potential points. finally, 28 wells had the lowest RMSE of 0.113 and had the best distribution. Location of wells obtained by a number of existing wells was near. The results obtained with the criteria of the quantity of groundwater monitoring network design, which corresponded to show the effectiveness of this approach.

In recent decades, Meta-heuristic algorithms has played an effective role in solving different engineering problems such as optimal operation of reservoirs. Owing to the complexity of water resources management problems, as well as, the daily growing need for the development and expansion of these methods, in this research, a model based on Symbiotic Organisms Search (SOS) algorithm was developed for modeling the optimal operation of complex multi-reservoirs systems. In the first step, the performance of the method was successfully assessed through several benchmark functions. Then it was used for the monthly operation of Tangemashure, Sazbon and Karkheh reservoirs located in Karkheh basin. The optimal allocation were considered for meeting the irrigation demands of 4 agricultural regions, and priority was with allocation of water for the environmental demands., for a 5 year period (from 1980-81 to 1984-85). The results of SOS algorithm were compared with other developed evolutionary algorithms including Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). The results indicated that, optimized operation policy through SOS algorithm with the sustainability index of 99.99, 99.11, 82.92 and 79.47 percent for Sazbon, Tangemashure, upstream and downstream of Karkheh reservoirs was more appropriate performance as compared to GA and PSO algorithms in optimal operation of multi-reservoirs systems.