Information Technology Research Base of Civil Aviation Administration of China

Tianjin, China

Information Technology Research Base of Civil Aviation Administration of China

Tianjin, China
SEARCH FILTERS
Time filter
Source Type

Guo W.,Nanjing University of Aeronautics and Astronautics | Guo W.,Yancheng Teachers University | Xu T.,Nanjing University of Aeronautics and Astronautics | Xu T.,Civil Aviation University of China | And 2 more authors.
Neural Computing and Applications | Year: 2016

An M-estimator-based online sequential extreme learning machine (M-OSELM) is proposed to predict chaotic time series with outliers. The M-OSELM develops from the online sequential extreme learning machine (OSELM) algorithm and retains the same excellent sequential learning ability as OSELM, but replaces the conventional least-squares cost function with a robust M-estimator-based cost function to enhance the robustness of the model to outliers. By minimizing the M-estimator-based cost function, the possible outliers are prevented from entering the model’s output weights updating scheme. Meanwhile, in the sequential learning process of M-OSELM, a sequential parameter estimation approach based on error sliding window is introduced to estimate the threshold value of the M-estimator function for online outlier detection. Thanks to the built-in median operation and sliding window strategy, this approach is efficient to provide a stable estimator continuously without high computational costs, and then the potential outliers can be effectively detected. Simulation results show that the proposed M-OSELM has an excellent immunity to outliers and can always achieve better performance than its counterparts for prediction of chaotic time series when the training dataset contains outliers, ensuring at the same time all benefits of an online sequential approach. © 2016 The Natural Computing Applications Forum


Xu T.,Civil Aviation University of China | Xu T.,Information Technology Research Base of Civil Aviation Administration of China | Meng Y.,Civil Aviation University of China | Lu M.,Civil Aviation University of China | Lu M.,Information Technology Research Base of Civil Aviation Administration of China
Dianzi Yu Xinxi Xuebao/Journal of Electronics and Information Technology | Year: 2016

Process mining is a technology which can extract non-trivial and useful information from airport event logs. However, the airport event logs are always on a detailed level of abstraction, which may not be in line with the expected abstract level of an analyst. Process models generated by these event logs are always spaghetti-like and too hard to comprehend. An approach to overcome this issue is to group low-level events into clusters, which represent the execution of a higher-level activity in the process model. Therefore, this paper presents a new activity mining method which is based on RankClus algorithm to generate activity clusters integrated with ranking. On this basis, the activity-clustered model which is easier to comprehend can be constructed. The experiment results show that this activity-clustered model, which shares a similar level of conformance with the meta model, is significantly less complex. © 2016, Science Press. All right reserved.


Xu T.,Civil Aviation University of China | Xu T.,Information Technology Research Base of Civil Aviation Administration of China | Yang Q.-C.,Civil Aviation University of China | Lu Z.-L.,Civil Aviation University of China | Lu Z.-L.,Information Technology Research Base of Civil Aviation Administration of China
Dianzi Yu Xinxi Xuebao/Journal of Electronics and Information Technology | Year: 2014

The prediction of airport noise plays an important role in airport noise control, flight schedule planning and surrounding designs of airport. However, the existing prediction models are complex and need so many highly accurate parameters that are monitored and collected as input of the model, hence adding difficulties to the prediction of airport noise. In order to solve these problems, this paper presents a prediction model based on the rough set and ensemble learning. Accordingly, the attributes of monitored noise data around airport is first reduced by the rough set and the subsets of attributes is produced then, the dynamic ensemble learning is used to combine base learners which are presented in three-dimensional coordinates based on the subsets of attributes. The results of experiments show that the proposed model can predict the noise of specific aircraft with full parameters being more accurately than existing models. And even if there is a lack in part of parameters, the prediction outcome of the model is able to approach the real value of airport noise while gradually increasing parameters.


Tang X.,Tianjin University | Xu T.,Civil Aviation University of China | Xu T.,Information Technology Research Base of Civil Aviation Administration of China | Feng X.,Tianjin University | And 2 more authors.
PLoS ONE | Year: 2014

Uncovering community structures is important for understanding networks. Currently, several nonnegative matrix factorization algorithms have been proposed for discovering community structure in complex networks. However, these algorithms exhibit some drawbacks, such as unstable results and inefficient running times. In view of the problems, a novel approach that utilizes an initialized Bayesian nonnegative matrix factorization model for determining community membership is proposed. First, based on singular value decomposition, we obtain simple initialized matrix factorizations from approximate decompositions of the complex network's adjacency matrix. Then, within a few iterations, the final matrix factorizations are achieved by the Bayesian nonnegative matrix factorization method with the initialized matrix factorizations. Thus, the network's community structure can be determined by judging the classification of nodes with a final matrix factor. Experimental results show that the proposed method is highly accurate and offers competitive performance to that of the state-of-the-art methods even though it is not designed for the purpose of modularity maximization. Copyright: © 2014 Tang et al.


Xu T.,Civil Aviation University of China | Xu T.,Information Technology Research Base of Civil Aviation Administration of China | Xie J.,Civil Aviation University of China | Yang G.,Information Technology Research Base of Civil Aviation Administration of China
Nanjing Hangkong Hangtian Daxue Xuebao/Journal of Nanjing University of Aeronautics and Astronautics | Year: 2013

According to the characteristics of airport noise data, a fast hierarchical clustering algorithm based on representative point is presented. This algorithm improves the traditional condensed hierarchical clustering algorithm by using clustering representative point method and dichotomy strategy. Meanwhile, a clustering result evaluation method which combines the clustering representative point and the definition of similarity in clustering algorithm is proposed. The experimental results show that the proposed algorithm not only has high performance, but also can discover the noise distribution model of a specific type of flight event correctly. The method can accuratly predict the noise distribution model of these flight events.


Xu T.,Civil Aviation University of China | Xu T.,Information Technology Research Base of Civil Aviation Administration of China | Su H.,Civil Aviation University of China | Yang G.-Q.,Information Technology Research Base of Civil Aviation Administration of China
Zhongguo Huanjing Kexue/China Environmental Science | Year: 2016

This paper proposes an airport noise ensemble prediction model based on space fitting and neural network by introducing ensemble learning method. Space fitting and BP neural network is used respectively to create the base learner and a heterogeneous ensemble algorithm based on observational learning is used to integrate these base learners. The final ensemble model thus can improve prediction accuracy effectively by integrating multiple heterogeneous base prediction learners. The experimental results shows that the proposed heterogeneous ensemble algorithm based on observational learning is better than other heterogeneous ensemble algorithms on accuracy and tolerance for solving the airport noise prediction problem. © 2016, Chinese Society for Environmental Sciences. All right reserved.


Xu T.,Civil Aviation University of China | Xu T.,Information Technology Research Base of Civil Aviation Administration of China | Yan X.-J.,Civil Aviation University of China | Yang G.-Q.,Information Technology Research Base of Civil Aviation Administration of China
Zhongguo Huanjing Kexue/China Environmental Science | Year: 2014

Through analyzing the influence factors of the noise event for single flight, the regression prediction model based on BP neural network was established. Then, the ensemble prediction model based on neural network for single noise event was constructed by selecting neural networks with the aid of adaptive genetic algorithm. Simultaneously, in order to maintain the diversity of neural networks, different number of hidden nodes and Bagging algorithm were used. Experimental results show that the proposed ensemble prediction model based on neural network was better than the model of single BP neural network in terms of generalization ability and higher stability. The average accuracy rate of the proposed model was 96.9% on the test set within ±3 dB error and was 6.8% higher than that of the single network model.


Zou P.-C.,Nanjing University of Aeronautics and Astronautics | Wang J.-D.,Nanjing University of Aeronautics and Astronautics | Yang G.-Q.,Information Technology Research Base of Civil Aviation Administration of China | Zhang X.,Nanjing University of Aeronautics and Astronautics | Wang L.-N.,Nanjing University of Aeronautics and Astronautics
Ruan Jian Xue Bao/Journal of Software | Year: 2013

An effective distance metric is essential for time series clustering. To improve the performance of time series clustering, various methods of metric learning can be applied to generate a proper distance metric from the data. However, the existing metric learning methods overlook the characteristics of time series. And for time series, it is difficult to obtain side information, such as pairwise constraints, for metric learning. In this paper, a method for distance metric learning based on side information autogeneration for time series (SIADML) is proposed. In this method, dynamic time warping (DTW) distance is used to measure the similarity between two time series and generate pairwise constraints automatically. The metric which is learned from the pairwise constraints can preserve the neighbor relationship of time series as much as possible. Experimental results on benchmark datasets demonstrate that the proposed method can effectively improve the performance for time series clustering. ©Copyright 2013, Institute of Software, the Chinese Academy of Sciences.


Chen H.,Nanjing University of Aeronautics and Astronautics | Chen H.,Information Technology Research Base of Civil Aviation Administration of China | Yang B.,Nanjing University of Aeronautics and Astronautics | Xu T.,Information Technology Research Base of Civil Aviation Administration of China | Wang J.,Nanjing University of Aeronautics and Astronautics
Nanjing Hangkong Hangtian Daxue Xuebao/Journal of Nanjing University of Aeronautics and Astronautics | Year: 2013

To predict and prevent the noise around the airport becomes an urgent problem. A new airport noise prediction model based on fuzzy support vector regression is established on the existing generic software for noise calculation. To integrate into the fuzzy support vector regression algorithm, the fuzzy membership of each sample is determined by its local outlier factor. Finally, the prediction accuracy, noise immunity, generalization ability of the proposed model are validated on the historic flight data of an airport. Experiments show that the fuzzy support vector regression algorithm based on local outlier factor can effectively predict the noise levels around airports, and is more accurate and better noise immunity than the standard support vector regression.


Lu M.,Civil Aviation University of China | Lu M.,Information Technology Research Base of Civil Aviation Administration of China | Feng X.,Civil Aviation University of China | Feng X.,Information Technology Research Base of Civil Aviation Administration of China
Journal of Computational and Theoretical Nanoscience | Year: 2015

The training and testing data in many machine learning problems naturally falls into distinct groups. For example, in email spam filtering the emails are naturally grouped by email accounts; in handwritten character recognition the handwritten characters are naturally divided into groups by hand writers. The data distributions of these distinct groups are different, which is a prior knowledge for learning models. However, the traditional machine learning methods ignore such prior knowledge, which exerts adverse effect on their performance in real-world applications. To address this issue, the paper learns a group-specific model for each group in training, and then predicts data from unseen group by the group-specific model created from one nearest neighbor group of the unseen group. These group-specific models are learned from not only the group itself but also with the assistance of other groups through a group correlation matrix. Due to the highly coupling between model parameters and group correlation matrix in the objective function, an alternating method is employed to learn them. In addition, a novel data projection method is developed to find the nearest neighbor group from training data for a given unseen group. Experimental results on two text applications show the advantage of the proposed algorithm. © 2015 American Scientific Publishers.

Loading Information Technology Research Base of Civil Aviation Administration of China collaborators
Loading Information Technology Research Base of Civil Aviation Administration of China collaborators