Multi-Label classification aims at assigning more than one Label to each instance. Many real-world multi-Label classification tasks are high dimensional, leading to reduced performance of traditional classifiers. Feature selection is a common approach to tackle this issue by choosing prominent features. Multi-Label feature selection is an NP-hard approach, and so far, some swarm intelligence-based strategies and have been proposed to find a near optimal solution within a reasonable time. In this paper, a hybrid intelligence Algorithm based on the binary Algorithm of particle swarm optimization and a novel local search strategy has been proposed to select a set of prominent features. To this aim, features are divided into two categories based on the extension rate and the relationship between the output and the local search strategy to increase the convergence speed. The first group features have more similarity to class and less similarity to other features, and the second is redundant and less relevant features. Accordingly, a local operator is added to the particle swarm optimization Algorithm to reduce redundant features and keep relevant ones among each solution. The aim of this operator leads to enhance the convergence speed of the proposed Algorithm compared to other Algorithms presented in this field. Evaluation of the proposed solution and the proposed statistical test shows that the proposed approach improves different classification criteria of multi-Label classification and outperforms other methods in most cases. Also in cases where achieving higher accuracy is more important than time, it is more appropriate to use this method.

Detection of overlapping communities in large complex social networks with intelligent agents, is an NP problem with great time complexity and large memory usage and no simultaneous online solution. Proposing a novel distributed Label Propagation approach can help to decrease the searching time and reduce the memory space usage. This paper presents a scalable distributed overlapping community detection approach based on the Label Propagation method by proposing a novel Algorithm and three new metrics to expand scalability and improve modularity through agent-based implementation and good memory allocation in a multi-core architecture. The experimental results of large real datasets over the state-of-the-art SLPA approach show that the execution time speeds up by 900% and the modularity improves by 3% to 100% thus producing fast and accurate detection of overlapped communities.

Detecting communities in large networks is an important challenge in social network analysis, and providing an Algorithm with optimal accuracy and efficiency for extracting communities is very important. There are different approaches to identify communities in social networks, including methods based on classical clustering, Algorithms based on criteria of similarity in features, methods based on finding subgraphs with a lot of internal communication, as well as a Label Propagation approach. In the Label Propagation approach, first, the most important vertices of the network are determined based on a series of importance and centrality criteria, and different community Labels are assigned to them. Then the Label of each of these vertices is propagated to the neighboring vertices and around them. The aim of this research is to improve a community detection Algorithm called LBLD. This Algorithm first determines five percent of the most important network vertices based on a similarity criterion. Then, with a balanced approach, communities are developed both from the center and from the borders, and finally a phase of integration is implemented so that small communities are combined with each other and form desirable communities. Our proposed idea uses a measure inspired by the concept of h-index to improve the accuracy of community detection. In such a way that subgraphs are identified as communities that have at least p percent of vertices with at least degree k. The accuracy and efficiency of the proposed solution have been evaluated by applying it to known data sets in this field and it shows a significant improvement compared to existing similar methods.

Multi-Label learning Algorithms face many challenges due to the high volume and dimensions of multi-Label data and the existence of noise. Feature selection methods are an effective technique for addressing these challenges. This paper presents a feature selection method based on an ensemble approach for multi-Label data. In this approach, three different decision matrices based on various feature evaluation criteria, taking into account the relevancy of features with class Labels and their redundancy relative to each other, are effective in the feature selection process. These three decision matrices are finally combined based on an ensemble approach using the concept of fuzzy integral to evaluate the features according to the aggregate value. Comparisons have been made with several similar Algorithms to illustrate the performance of the proposed method.

Community detection in dynamic social networks is one of the most important research topics that has been considered in recent years. There are various approaches to detecting communities in dynamic social networks, among which the Label Propagation approach has chosen due to simplicity and efficiency. This approach involves many methods that are often random. Among these methods, LabelRankT(Time) is a deterministic method. Of course, this method also has problems, one of the problems is that when a node wants to join a community, the internal structure of that community is not considered. So, for solving this problem, a greedy approach has been used to improve the Label publishing approach. The proposed approach and other evaluated methods are tested in real and Synthetic datasets. The results show that the proposed method has performed relatively better than the other methods in terms of accuracy and modularity.

The use of neural networks methodology is not as common in the investigation and prediction noise as statistical analysis. The application of artificial neural networks for prediction of power tiller noise is set out in the present paper. The sound pressure signals for noise analysis were obtained in a field experiment using a 13-hp power tiller. During measurement and recording of the sound pressure signals of the power tiller, the engine speeds and gear ratios were varied to cover the most normal range of the power tiller operation in transportation conditions for the asphalt, dirt rural roads, and grassland. Signals recorded in the time domain were converted to the frequency domain with the help of a specially developed Fast Fourier Transform (FFT) program. The narrow band signals were further processed to obtain overall sound pressure levels in A-weighting. Altogether, 48 patterns were generated for training and evaluation of artificial neural networks. Artificial neural networks were designed based on three neurons in the input layer and one neuron in the output layer. The results showed that multi layer perceptron networks with a training Algorithm of back Propagation were best for accurate prediction of power tiller overall noise. The minimum RMSE and R2 for the four-layer perceptron network with a sigmoid activation function, Extended Delta-Bar-Delta (Ext. DBD) learning rule with three neurons in the first hidden layer and two neurons in the second hidden layer, were 0.0198 and 0.992, respectively.

uspended particles have deleterious effects on human health and one of the reasons why Tehran is effected is its geographically location of air pollution. One of the most important ways to reduce air pollution is to predict the concentration of pollutants. This paper proposed a hybrid method to predict the air pollution in Tehran based on particulate matter less than 10 microns (PM10), and the information and data of Aghdasiyeh Weather Quality Control Station and Mehrabad Weather Station from 2007 to 2013. Generally, 11 inputs have been inserted to the model, to predict the daily concentration of PM10. For this purpose, Artificial Neural Network with Back Propagation (BP) with a middle layer and sigmoid activation function and its hybrid with Genetic Algorithm (BP-GA) were used and ultimately the performance of the proposed method was compared with basic Artificial Neural Networks along with (BP) Based on the criteria of - R2-, RMSE and MAE. The finding shows that BP-GA R2 = 0.54889 has higher accuracy and performance. In addition, it was also found that the results are more accurate for shorter time periods and this is because the large fluctuation of data in long-term returns negative effect on network performance. Also, unregistered data have negative effect on predictions. Microsoft Excel and Matlab 2013 conducted the simulations.