JOURNAL OF SOFT COMPUTING AND INFORMATION TECHNOLOGY (JSCIT)
Year:2019 | Volume:8 | Issue:3 |Papers Count:7

Detecting social groups is one of important and complex problems which has been concerned recently. Detecting social groups and relation between group members will be necessary for human robots in near future. Databases have some information including trajectories and also labels of members. The target is to detect social groups that contains at least two people or detecting individual motion of the people. In the proposed method, for detecting social groups, physical distance, temporal causality and shape similarity features are used. The required time to extract these features is lower than the other suggested features. In addition to precession and recall, the effectiveness of the proposed method in terms of required time for training and testing data is also examined. Lower required time provides greater ability to implement for human robots. The proposed method provides acceptable results in valid databases and is compared to existing methods in terms of statistical results and the required time.

Gravitational Search Algorithm (GSA) is a simple and efficient optimization method recently proposed for solving single-objective optimization problems. In this paper, for the first time, the nearest-better neighborhoods are defined in swarm intelligence algorithms and then used in the GSA to solve multi-modal optimization problems. For this purpose, two neighborhoods are defined, called Topological Nearest-Better (TNB) and Distance-based Nearest-Better (DNB), and then these two structures are used separately in the GSA and two different versions of the GSA for multi-modal optimization problems are provided. To investigate the efficiency of the proposed algorithms, an empirical assessment has been performed on several standard multi-modal benchmark functions. The results of these experiments show that the proposed algorithms can achieve good results compared to other multi-modal optimizer algorithms.

Wireless sensor networks include low-weight low cost nodes equipped with limited processing and energy sources. Sensor nodes monitor their surroundings and send sensed events to the sink node through shortest path. One of the major challenges in these networks is to keep nodes connected to each other while the target area is covered effectively. These two parameters are referred as main parameters of quality of service (QoS). In this paper, optimal area coverage is carried out using the Gravitational Search Algorithm. In the proposed algorithm, the agents or masses are equivalent to sensor nodes. The nodes are influenced by distance, Newton gravity law and the laws of motion. A group of nodes can be considered as a cluster. The proposed algorithm is compared with previous methods in terms of network life time, remaining energy and network power. Simulation results show that the proposed method reduces energy consumption by optimizing the number of nodes in the area with maximal coverage, and increases the lifetime and effectiveness of the network.

Measurement and experimental analysis of stress and strain of enforced objects are crucial in the fields of mechanics and civil engineering. The photoelasticity as a conventional method for measurement and analysis suffers from some limitations such as the need for specific transparent material, appropriate equipment and enough experience. In this research, photogrammetry was introduced for the experimental analysis of stress and strain measurement because of its high accuracy, ease and independence to the material of the object. To compare the accuracy of photogrammetry and photoelasticity, a crane hook-shaped object from Araldite epoxy was tested for different enforcement. In each step, in addition to recording the required information for photoelasticity, photos of the object were taken by a digital camera in a fixed position and orientation. The positions of the corresponding points on the object were measured with an accuracy of 0. 01 pixel using digital image processing and least square image matching techniques. The measured stress and strain using photoelasticity and photogrammetry were compared with analytical stress and strain measurement method. The results indicated high accuracy for photogrammetry compared to photoelasticity. Therefore, conventional methods for stress and strain measurements can be replaced by photogrammetry.

Dominant and rare events detection is one of the most important subjects of image and video analysis field. Due to inaccessibility to all rare events, detecting of them is a challenging task. Today, deep networks are the best tool for video modeling but due to inaccessibility to tagged data of rare data, usual learning of a deep convolutional network is not possible. Due to the success of generative adversarial networks, in this paper an end-to-end deep network based on generative adversarial networks is presented for detecting rare events. This network is competitively trained only by dominant events. To evaluate performance of proposed method, two standard datasets: UCSDped1 and UCSDped2 are utilized. The proposed method can detect rare event with 0. 2 and 0. 17 equal error rate with the processing speed of 300 frames per second on the mentioned data respectively. In addition to end-to-end structure of the network and its simple train and test phase, this result is comparable to advanced methods results.

The PageRank algorithm is one of the web-based classification methods used by Google search engine first. The main purpose of this algorithm was to determine the popularity of Web pages. The algorithm uses the web links structure to find important pages. One of the problems of PageRank and the same algorithms based on the web graph is that the number of a page is propagated to its output pages without any control while the output screen really is not really recommended by the previous page directly or indirectly. In this paper, by changing the original formula PageRank, a method has been proposed to prevent the entry of the input bonds to a single page of publication without gaining popularity. In order to evaluate the proposed algorithm, a single web graph is constructed, which in some nodes has a rating leap. This mutation decreases after the proposed algorithm is applied.

This research focuses on the usefulness of various intelligent machine learning algorithms on prediction of time series in financial markets. A challenge in this area is that economic managers and the scientific community are still demanding predictive algorithms with greater accuracy. The elimination of the mentioned challenge can improve the quality of the predictions and, as a result, lead to higher profitability and productivity. The proposed solution relies on finding the best input variables by using the regression-based machine learning algorithms, with emphasis on the leading selection methods. We implemented the concerned ideas using the Python language and the relevant machine learning tools. In our experiments, as dataset, we used the stock information of two companies from the Tehran Stock Exchange. These datasets belong to the transactions accomplished in years 2008 to 2018. The experimental results show that the technical features selected by the forward method can find the most effective and also the best values for the required parameters. The experimental results and formal analyses indicate that the use of selected technical features as inputs to the support-vector-machine and to the multi-layer perceptron machine gives prediction with the least-error, and this would provide more accurate predictions.