random forest algorithm

Selection of quarries and quality stones by design standards has always been one of the most important challenges in the location of breakwaters and the commencement of executive activities of rock mass breakwater construction projects. This research proposes using machine learning approaches to predict the result of the acceptance or rejection of quarry stones with the minimum possible tests based on the algorithm's output. The method used in this thesis is to provide a framework including complete and accurate data preprocessing and the use of machine learning algorithms such as decision tree, random forest, nearest neighbor, and the use of available data from the results of rock tests in the last ten years obtained in the construction of rock mass breakwaters in the coastal strip of the Sea of Oman. In this method, the available data is classified into two parts: main data and test data, and the algorithm is implemented on the main data and then the output of the algorithm is evaluated using the test data. The algorithm output is confirmed with an accuracy of 96%. The results obtained from this framework emphasize the importance of using existing data in the marine construction industry and the effectiveness of using machine learning algorithms in analyzing and interpreting existing data. The output of the results of this thesis reduces the time of experiments, reduces project costs, and reduces the duration of the project. The insight obtained from this research can help companies active in the field of marine construction and also specifically the Ports and Maritime Organization as the custodian of the construction and maintenance of marine structures in the country to locate breakwaters, optimize resource allocation, reduce the implementation time and operation of projects.

The development of metasurface absorbers offers a potential solution to achieve low weight, thin thickness, favorable absorption rate, and acceptable terahertz absorption characteristics. In order to optimize the absorption properties of metasurfaces, the absorption spectrum is usually used as an important evaluation criterion, which can show many important characteristics such as the rate of absorption at different frequencies. However, the analysis of absorption spectra related to a large number of variable structural parameters is required when designing the structure, which consumes a lot of resources and time, because electromagnetic wave absorption involves the processes of complex impedance matching and electric field excitation. To address this issue, this study proposes a machine learning approach based on a random forest algorithm to predict absorption rates based on structural parameters, reducing the need for numerical simulation and spectrum analysis time. With the random forest model, the absorption rate is predicted with the R2 score of more than 0.99. In addition, the proposed absorber design has the advantages of being thin, insensitive to polarization and with a relatively stable incident angle, due to the symmetry of the structure. This study presents a practical and effective approach for the design of complex systems related to absorbing, reflecting and transmitting electromagnetic wave propagation.

The purpose of this research is to investigate the amount of relative energy loss (EDR) in zigzag weirs with triangular and trapezoidal plans in different dimensions using Support Vector Machine (SVM) model, Random Forest (RF) algorithm, and Artificial Neural Network (ANN) method. 70% of the experimental data sets were used for the training phase and 30% for the test phase. In the SVM model, the results of different kernels showed that the Radial Basis Function (RBF) kernel has better results in predicting the relative energy loss of zigzag weirs compared to the Polynomial, Linear, and Sigmoid kernels. The results of statistical indicators of correlation coefficient (R), percentage Mean Relative Error (Mean RE%), Root Mean Square Error (RMSE), and Kling Gupta Efficiency (KGE) for the SVM-RBF model in the test phase are 0.907, 1.38%, 0.0153, and 0.744, respectively. In the ANN method, the Multi-Layer Perceptron (MLP) network has more accurate results compared to the RBF network. The results of the above indicators in the test phase for the ANN-MLP method are 0.969, 0.73%, 0.007, and 0.968, respectively. In addition, these results for the RF model are 0.878, 1.78%, 0.0192, and 0.362, respectively. Examining the results showed that the ANN method performs better than other SVM and RF models.

In recent years, the problem of maintaining the security of smart homes, where a large number of devices use Internet connections to communicate, has become one of the main concerns in the field of network security. Although a lot of research has been done to establish the security of smart homes, but considering the scope of the topic under discussion, most of these works do not have the necessary efficiency in terms of accuracy and speed of operation. In the proposed method, after performing some pre-processing operations on the dataset, with the help of Principal Component Analysis (PCA), a subset of the features of the dataset are selected to prepare the data for classification, which are the most effective features in intrusion detection. It is considered that this action leads to an increase in the accuracy and speed of the classification action. Also, in the classification stage, the random forest algorithm, which is a powerful algorithm based on machine learning, has been used on a very new dataset of the Internet of Things, called IoTID20. The proposed approach has shown high performance for intrusion detection with an accuracy of 99.73% and 98.46% for the classification of binary and multi-class attacks. Comparing the results of the proposed method with other works, it shows the superiority of the proposed method in detecting multi-class attacks.

Customer churn is a financial term that refers to the loss of a customer; Today, due the large number of banks , the loss of customers from one bank to another has become a serious concern for different banks. Therefore, in this article, which has been compiled for the customers of a bank , it is possible to identify customers who have a high probability of falling by considering the behavior and characteristics of the customers before the fall occurs and to keep them by providing suggestions. In marketing, everyone agrees that keeping a customer is much less expensive than attracting a new customer, this article introduces the different phases of the approach of predicting customer churn with the help of machine learning. The proposed method is a combination of random forest algorithms and Jaya optimization, and customer dropout is based on different characteristics. Customer like age, Gender, graphs and cases It predicts more . The results of model in the article are 91.41%, 95.66% and 93.35% respectively in Precision , Recall and Accuracy criteria.

In this research, a new approach for image segmentation based on genetic algorithm and random forest is presented. Image segmentation can be done using supervised learning. In this learning, there are a number of images from data set with their labels. In image segmentation, different parts of the image separate from each other. In this process, all the pixels in the image are given a label, so that the pixels with the same label have common characteristics with each other. To provide a model that can perform image segmentation, it is necessary to extract features from input images and perform segmentation using a suitable classifier and these features. Image feature extraction is done using image filters. In this research, a hybrid combination of 4 Gabor filter banks and Sobel, Prewitt, Canny edge, Scharr, Gaussian, median, and Roberts filters are used for effective feature extraction. One of the most important of these filters, which also has a degree of freedom, is the Gabor filter. This filter has a number of   hyperparameters that change the efficiency of the classifier by changing these hyperparameters. In this research, an attempt has been made to adjust these hyperparameters using genetic algorithm. The fitness function proposed in this research is f1-score. random forest classifier is utilized for image segmentation and classification. The results of the experiments show that the hyperparameters found by the genetic algorithm have been able to perform a satisfactory segmentation on data set.

The growing number of malware is one of the major threats in the field of cyber and its detection has always been associated with challenges. Windows-based malicious executable files perform malicious activity at the target operating system level or any other application by manipulating features in their header and obscuring their behavior. Detecting suspicious specimens from a large volume of input samples as well as discovering new and unknown malware is one of the topics that is always researched by researchers. In this study, a combined method has been proposed to determine the level of maliciousness of suspicious executable files. Kashif's proposed method consists of two static modules for extracting executable file header properties, and two behavioral modules for extracting signature-generating properties and a thoughtful behavioral model based on machine learning methods. The purpose of this study is to identify suspicious Windows executable files from the large volume of files and determine their maliciousness level. This method detects malware based on the maliciousness probability of being assigned to each file. Experiments showed a malignancy percentage of six types of malware for PE header detector module, in the range of 62.7 to 70%, Yara-based detector module, in the range of 70.8 to 78.2%, Behavioral signature-based detector module, 98% and ML-based detector module by using Random forest learning algorithm has been 99% accuracy. The experimental results also showed that Kashef detected 94% of the protected malware with a 2% improvement compared to the results of 10 similar products. And with 98% detection of unprotected malware, there is a 5% improvement compared to the results of 10 similar products.

Despite all the efforts of security experts to detect SQL injection attacks, according to OWASP  report’s, SQL injection attack is still used as the most important cyber attack by attackers. In order to detect attacks, two methods are used: signature-based and behavior-based. Signature-based methods are used for known attacks, and behavior-based methods are suitable for detecting unknown attacks. Behavior-based intrusion detection systems are more useful because attacks are implemented in different ways. Behavior can be analyzed by methods such as classification, clustering, etc. One of the most important classification algorithms is the random forest algorithm which has high accuracy and on the other hand the implementation and interpretation of the results can be done easily using this algorithm. According to the studies, the accuracy of the random forest algorithm is highly dependent on its input parameters. These parameters include 9 items, including the number of trees, their depth, voting method, information gain, and so on. Optimal determination of these parameters is an optimization problem with large state space. In this research, a method based on genetic algorithm to determine the optimal values of these parameters is presented. Due to the optimal determination of the parameters, the obtained results show an improvement in the detection accuracy compared to the default state of the algorithm and other researches. The evaluation results indicate that the intrusion detection accuracy in the proposed method was %98, which is about %11 higher than the random forest algorithm with default parameters and %08 higher than previous studies.

Due to increasing population and decreasing energy sources, this research studies the consumption of domestic energy. The purpose of this study is to predict the factors affecting household energy consumption. To do this, we use 3 algorithms, M5Rules, K-nearest neighbor and random forest, available in Weka software. In this study, the feature correlation algorithm is used to select the most important factors affecting energy consumption and their impact. The results show that lights and fixtures, temperature of the living room, outside temperature, temperature outside of Chievres Station, wind speed, humidity in the kitchen and the temperature in the laundry area have the most impact on household energy consumption. Among the methods, random forest algorithm presented the best results.

As computer networks grow, so attacks and intrusions to these networks are increased. In order to have a fully secure computer network, one needs ‘intrusion detection systems’ (IDS) on top of firewalls. The goal of using an IDS is to supervise the abnormal activities and differentiate between normal and abnormal activities in a host system or in a network. An efficient IDS has high detection rate while keeping a low false alarm rate. In this paper, a new approach to classify KDD-Cup-99 data set using a combination of random forest method and genetic algorithm is presented. The purpose is to increase the speed of learning and test phases while improving the accuracy. Random forest is an ensemble learning method based on decision trees. Due to its relatively simple structure and good performance, it is used in many supervised learning applications.  However, like all tree based machine learning algorithms, having too many categorical features, can be a problem both for the speed and accuracy. This is exactly the case with the problem in hand, i.e. intrusion detection; many of the features are in the form of categorical data. For example, in R language, the maximum number of definable categorical features for random forest is 53. The contribution of this work is resolving this issue with the aid of Genetic Algorithm (GA). In this research information gain as a measure of importance is defined and the number of features is reduced based on genetic algorithm.

This paper presents a novel method based on machine learning strategies for fault locating in high voltage direct current (HVDC) transmission lines. In the proposed fault-location method، only post-fault voltage signals measured at one terminal are used for feature extraction. In this paper، due to high dimension of input feature vectors، two different estimators including the generalized regression neural network (GRNN) and the random forest (RF) algorithm are examined to find the relation between the features and the fault location. The results of evaluation using training and test patterns obtained by simulating various fault types in a long overhead transmission line with different fault locations، fault resistance and pre-fault current values have indicated the efficiency and the acceptable accuracy of the proposed approach.