ant colony algorithm

In this study, the Quantitative Structure-Property Relationship (QSPR) was proposed to predict the Upper Flammability Level (UFL) of 588 organic compounds including hydrocarbon compounds, halogenated compounds, alcohols, ethers, esters, aldehydes, ketones, acids, amines, amides, nitriles, and nitro compounds. A variety of molecular descriptors were calculated for each molecule. The Memorized-Ant Colony Algorithm (M-ACO) combined with multivariate linear regression (MLR) was used to select the best subset of descriptors that have a significant contribution to the UFL property. Different variable transformations were performed on both dependent and independent variables to obtain better multiple linear regression models. The best model was a four-variable model obtained by using the calculated descriptors as independent variables and the logarithm of UFL values as the dependent variable. This model has a very wide applicability range of UFL from 2/7 to 100 vol %. The training and test errors of the model were found to be 0/1 log UFL unit (R2 = 0.80) and 0.12 log UFL unit (R2 = 0.75), respectively. Therefore, the model has good accuracy and can be used to predict the UFL of a wide range of organic compounds.

Evaluation of Longshore Sediment Transport Rate in Imam Khomeini port using three examples of meta-heuristics algorithmImam Khomeini port is one of the ports of Khuzestan province in the Persian Gulf, which has 38 active wharves with a length of 7 km, is the largest active port in Iran and the annual dredging volume in this port is about 600 thousand cubic meters, of which about 400 thousand cubic meters is related to the Duragh canal and about 200 thousand are related to other ports of the port, since the sedimentation rate and dredging cost in this port is significant, so, in this study the rate and volume of sedimentation based on field information and modeling obtained from the Ports and Maritime Organization And comparing these values with the results obtained from meta-heuristic algorithms under MATLAB software in two-dimensional flow conditions and also the presence of all parameters affecting the transfer of sediment along the coast in Imam Khomeini port. After combining the three equations of CERK, Kamphuis and Bayram (CKB) and calculating the optimal coefficients and capabilities by Differential Evolution, Ant Colony Optimization and Genetic Algorithm, evaluation and validation with the values obtained from the available field data, coefficient error and computational capabilities Calculated using the root mean squares and bias method, therefore, it was observed that the coefficients and powers obtained from the genetic algorithm that were placed in the CKB equation have the most overlap with the rate longshore sediment transport from field data.

Inversion of basement relief of sedimentary basins is an important application among the non-linear modeling techniques. Particularly in sedimentary basins with hydrocarbon source potential, the thickness of sediments is one of the primary factors in determining the thermal maturation of these basins. Gravity methods have been vastly used to estimate the base of sedimentary basins. The aim of this research is two-dimensional modeling of the basement geometry of a sedimentary basin using the inversion of the gravimetry data. A common way to approach this problem is discretizing the basin using polygons (or other geometries), and solving the non-linear inverse problem by local optimization iteratively. This procedure provides a solution which highly depends on the initial model and the used prior information. Besides, due to the non-linearity of this inverse problem, local optimization methods will fail whenever there is no reliable initial model. The global optimization method is a promising alternative to classical inversion methods because the quality of their solutions does not depend on the initial model. Also, they do not use the derivatives of the objective function.Ant colony algorithm (ACO) is one of the kinds of important swarm intelligence algorithms which have been successfully applied in many fields such as inversion of geophysical data. This research, in two separate stages, investigates the design and implementation of the ACO as a powerful tool for two-dimensional non-linear modeling of gravity data. ACO can be a substitution for the local response methods such as Marquardt-Levenberg and Gauss-Newton. To apply this algorithm in the problem under consideration, it was validated with the data obtained from a synthetic model and then, reverse modeling of the real data was performed. For evaluating the validation of this developed algorithm, it was tested by the synthetic model. Data from the synthetic models were modeled by using the developed algorithm, and acceptable results were obtained. By using this approach, the topography of the basement in the synthetic model was obtained with acceptable accuracy. In this study, the effect of ACO algorithm on different values of probable noises was investigated. The results indicate that this algorithm is suitably stable against the Gaussian white noise with relatively high amplitudes. In modeling for high noise percentage, the root mean square error of the data calculated with the original data didn't exceed 1.64 mGal and that obtained with the original model at most was 131.4 m. The results of modeling show acceptable agreement with the original model even in the case of data contaminated with 10% Gaussian white noise.The reliability of the proposed method to the inversion of a real gravity data was confirmed by applying it on a real gravity profile in the Moghan sedimentary basin. The results of this modeling are compatible with previously published works in this area.

Staging is one of the most important factors determining the survival of a patient suffering from breast cancer and plays a key role in choosing treatment modalities. The method of choice for cancer staging is surgery followed by histological evaluation. However, finding a predictive algorithm to replace surgery in the staging of breast cancer is both cost- and time-saving and helps physicians to provide appropriate therapeutic techniques. The present paper introduces a strong predictive model for breast cancer staging using data mining techniques. We suggested a mechanized model based on the TNM staging system and the ant colony algorithm. This method would reduce the patient’s mental stress and financial costs because it does not need a surgical operation. The SEER international dataset and a local data set of 1148 women with breast cancer were used to evaluate the system performance, and model accuracy and the area under the ROC curve were calculated for different classifications. Using the TNM system, the accuracy rates were 99.93% and 99.91% for the SEER international dataset and the local dataset, respectively. The accuracy rates were 99.43% and 98.95% for the SEER international dataset and local dataset, respectively, when the ant colony algorithm was applied. Our results indicated that in addition to commonly used features in the TNM system, other features such as vascular invasion, age, blood group, number of children, birthplace, histology, CS Extension, positive regional node, morphology, and Site-Specific Factors 2, 3, and 6 can be used as important factors for breast cancer staging. Based on the results obtained, the two Logistic and Multi-Class Classifiers have the best accuracies for the SEER and local datasets in this study.

The objective of this study is the 3D modeling inversion of gravity anomalies in order to obtain a density model in different depth sections. We used a new method based on stochastic methods. Among the stochastic methods that can be applied for solving inverse optimization problems in geophysics are meta-heuristic algorithms which are based on artificial intelligence. The ant colony algorithm is categorized in this group of algorithms. It works on the basis of probability and trial and error and follows ant's behavior in finding the shortest distance between the nest and the food. This behavior of ants is closely similar to the inverse problems in geophysics which try to find the best solution for the unknowns in observation model. Therefore, this idea is applied for solving linear inverse problems. A MATLAB-based inversion code for the presented method was prepared. To examine the performance of this method, three different artificial models were assayed. The structure of these models was considered as a combination of 3- dimentional cubes so as to model every unknown geometrical structure. In the first example, our purpose was to investigate interferential anomalies resulting from two simple models with different density contrasts located in different depths. In the second example, to show the ability of the algorithm in an inversion using small anomalies, a model of an irregular geometry was assessed in different depths. Finally, in the third example, an interferential anomaly resulting from two models of complicated geometry, namely T and L and of different density contrasts was assessed. This method was applied for artificial models with and without noise. The results show that for an inversion by the use of the ant colony algorithm, there is no need to separate the interferential anomaly and it is possible to use it for a combination of density contrasts. Also, this algorithm is able to inverse anomalies of an order of 103MGAL. These anomalies belong to very small causative bodies such as: cavities and small orebodies. These anomalies are the main object of enviromental or engineering geophysics. The algorithm is semi-authomatic and search the best results without comprehensive pre-conditions. The method is well designed to consider the multiple anomalies in complex conditions. This character enables us to use it for interpretation of complex anomalies caused by geological sources where most of the semi-authomatic methods are useless. On the other hand, the inversion algorithm can be applied for different density contrasts. Relative positive or negative anomalies could be obtained by applying this method. This means that all different anomalies regarding their density contrasts can be detected and modelled through this method. Therefore we are not forced to isolate the object anomalies for inversion and this makes its application fast and easy for the whole surface of the Bougure residual anomalies. This character is very rarely obtainable in the published inversion algorithms. This character is particularly helpful when we would like to invert the precise data in the case of engineering geophysics. Another advantage of this method is its velocity due to its usage of probability and trial and error theory. This advantage is very important when we are facing with a large set of data and parameters to invert. The high number of parameters is a major problem in linear inversion methods, but can be treated by this method very properly.