Within vertical differentiation model, the role of targeted advertising on direct sales channel and distribution channel are studied. Compared to mass advertising, the firm using targeted advertising in distinct channels can benefit the firm's equilibrium profit by reducing the waste of advertising and enhancing the price of product. The manufacturer might use the targeted advertising to control the total profit of the channel. Meanwhile, the targeted advertising might reduce the retailer's profit as well as the total channel profit.
Older adults are increasingly using Social Networks Sites to support their social interactions. Moreover, the popularity of such network sites, the availability of datasets and recent progress in the computing systems and machine learning areas have made social network analysis a current important area of research. In this context, our current research aim is to investigate how these technological platforms are affecting the lifestyle of older adults. In this paper, we propose a Java oriented framework that can assist researchers in this area in the analysis of social networks groups, specifically in the comparison of user's groups creation from existing cluster-based algorithms. To validate the proposed framework we use a dataset extracted from the Meetup social network – a website providing members software services to schedule events using a common platform. For our study, we filtered the data for the specific group of older adults. The framework proposes several ways of evaluating the quality of the data, and is extensible to other clustering algorithms and evaluation metrics. Currently, we have tested our framework with the following well known clustering algorithms: k-means, fuzzy k-means, and affinity propagation. We report some preliminary results obtained by using the proposed framework and the above clustering algorithms using the extracted Meetup dataset.
This paper presents a case study, which describes a novel data collection and monitoring mechanism integrated into a food manufacturing company for the first time since they started their production over a century. Our results demonstrate this methodology provides a reliable approach to achieve a more consistent production facility. It also helps the factory understand the process and the parameters (e.g. oven temperature and environmental conditions) that significantly affect the product quality and consistency between production lines. We believe our results are make a useful contribution to the methodologies enabling intelligent manufacturing.
The correlation and treatment of lung cancer have been the focus of biomedical research. To establish a semantic knowledge network of lung cancer, we adopted the following strategies: firstly, lung-cancer related data was effectively fused into the existing knowledge base. Then, the RDF triple was used to organize and describe the data. The semantic knowledge network was of great significance because it could be used to identify the genes, proteins, drugs, and other factors related to the disease. All these factors are very useful in the diagnosis, preconditioning, and treatment of lung cancer. In order to solve the big data problem, a distributed parallel framework of PageRank algorithm was used to identify the key data in lung cancer pathway. The experimental results indicate the efficacy of this novel method.
In financial markets, the presence of inflation may affect investments' real returns. In this paper, we extend the fuzzy portfolio models to take into account the effects of inflation within the framework of credibility theory. The above models were deduced under the assumptions that returns of assets and the inflation rates follow some special fuzzy distributions. In a numerical study, we analyze the optimal portfolios and efficient frontiers with different inflation rates. We conclude that the absolute deviation increases as inflation rate increases, as a result, investment risks will be underestimated if the effects of inflation were not considered in the models.
Proposed a feasible algorithm for local community discovery in temporal networks named TNLCD. TNLCD algorithm is designed based on an improved temporal network model and a local modularity definition of our own for temporal networks. Experimental results on real-world datasets prove that TNLCD algorithm is not sensitive to initial nodes selection strategy and capable of revealing rich detailed community structure information which make the algorithm works well both in theoretical researches and practical applications.
Malaria is the leading cause of death in many countries. Numerous studies have been carried out to introduce prevention mechanisms; but most methods employed are limited to mathematical modelling and analysis. Predicting the occurrence of malaria incidence and understanding the dynamics of transmission still remain two key challenges. In this paper, we have utilised two different computational techniques to address these issues. We have used machine learning methods and developed models for predicting the likelihood of malaria outbreak using the incidence data against climatic factors. The success of machine learning depends on the availability of reliable large data sets; but in most cases it is not possible to have reliable and complete data. Also, machine learning does not provide a good understanding of the transmission mechanism. Hence, we have used agent-based modelling approach to simulate the malaria dynamics using some parameters. The model developed has potential to emulate real scenarios for showing impact of the incubation period on malaria dynamics. Moreover, the model can also assist hospitals, public health officials and policy makers with near-real evidence on how malaria infection invading population so as to strategies for feasible intervention.
There are two key problems in the field station systems related to the maintenance of long-distance data transmission and the power needed for a longterm operation of the field devices. In order to solve these problems, a new-design field station system that adopts the Internet of things technology and low-power consumption technology is proposed to achieve data acquisition and transmission in the field. The aim of the proposed field station system is to help researchers to get the valuable data in the field, and to provide some assistance support for in-depth scientific research. The overall design and parameters of the proposed system are explained in detail. Moreover, the conducted test shows that the proposed system can achieve not only the long-term stable data acquisition and transmission in the field with low-power consumption but also visualization of the scientific data. The proposed system has high practical value and has been used in the soil monitoring.
The objective of this study was to evaluate the main reproductive aspects of the marine cyclopoid copepod Cyclopina sp under laboratory conditions and to design, to implement and to validate a framework for the development of decision system support based on fuzzy set theory using clusters and dynamic tables. To validate the proposed framework, a fuzzy inference system was developed with the aim to estimate reproductive performance (Average fertility – AF, Reproductive frequency – RF, and Reproductive events number – REN) of the marine copepod cyclopina sp submitted to different thermal water and pH values conditions and compared with other Artificial Neural Networks models. The results show that the determination coefficients (R2) for the three output variables for the Fuzzy Inference System – FIS were 1.0, 1.0, and 0.999, respectively. The mean values of the standard deviations were 0.037 eggs, 0.061 hours, and 0.027 reproductive events, respectively, representing mean percentage errors of 0.240, 0.181, and 1.039 %, showing a better accuracy than the ANN-based one. The proposed framework may provide an effective means to draw a pattern to the development of fuzzy systems.
According to the problems of the poor resistance to interference peak pulse arising from doppler fuze information processing circuit, which is composed of the analog devices, the paper raises the target signal extreme recognition algorithm and there are two kinds of target signal algorithm no interference and interference. In the method proposed herein, each point is compared with the average of the points in the front. Regarding signal amplitude and signal frequency, the efficiency is improved. The test results show that the target signal extreme recognition algorithm can completely remove the peak pulse interference part in the signal, and the method improves the anti-interference ability of doppler fuze and it has significant reference in the modern fuze digital transformation and improvement of the anti-interference ability.
In order to improve the accuracy and recall rate of term extraction results in the Chinese patent domain, approaching from the perspective of deep learning, with part-of-speech and dependency relationships as features, a patent domain term extration model (Bi-LSTM-CRF) was proposed by combining Conditional Random Fields (CRF) and bi-directional long short-term memory (Bi-LSTM) based on a multi-feature fusion. Based on the two explicit characteristics of part of speech and dependency, the double-layer bidirectional LSTM neural network was used to mine the temporal and semantic information in the data, which overcame the disadvantages of the traditional methods, such as weak generality and inability to capture the implicit information in the context as well as addressing the dependency relationship among the output tags through the CRF layer. Experimental results show that this deep learning method is effective in terms of domain term extraction.
Patent value evaluation is very important in the areas of patent transaction and patent financing. By integrating K-means clustering and C-F uncertainty reasoning model with sample data analytics, we propose a patent value evaluation algorithm to quantify patent value according to the aspects of patent content, patent applied ranges and patent research background. The process and implementation method of patent value evaluation algorithm are discussed, and the example to evaluate application value of patents in the area of railway industry in international patent database produced by the European Patent Office is introduced. The results can adequately verify effectiveness of our proposed algorithm.
In this paper, the probability risk assessment method and the probability risk assessment model are used to quantitatively analyze the security probabilities of space robots' target capture tasks. Based on the analysis of the process of capturing mission of the space robots, uncertainty analysis, importance degree and sensitivity results, objectively evaluate the weak link in the whole process of space robots target capture, and provide reference for the analysis of space robot in-orbit fault.
In the credit risk analysis of corporate bonds, multiple risk factors need to be considered comprehensively. These factors often exist in complex nonlinear systems. It is difficult to accurately model these system equations by conventional methods. In view of the complexity and uncertainty in credit analysis, this paper attempts to transform the simulation of complex nonlinear systems into classification and recognition of credit risk factor images, transform risk factors into unstructured data, to propose a credit risk modeling and analysis method based on convolutional neural network (CNN). Experiments show that the accuracy of credit risk identification in corporate bonds in multiple industries is over 80%.
Understanding and predicting water evaporation patterns and factors is a critical issue in places in the world where water is a rare commodity, tropical countries which have large surface of water and high temperatures, or in desert countries with very high temperatures almost all over the year such as the state of Kuwait. Understanding different patterns of evaporation rate process is necessary as well as important. Previous literatures mainly focus onto using deterministic approach in understanding such patterns and favoring approximating its behaviors using concrete estimated equations. One of the most promising areas of understanding stochastic patterns from data like this is machine learning. This paper uses real data of different related attributes collected from the environment of the state of Kuwait to build, train, and test an accurate behavioral model created by different machine learning algorithms to do prediction of such process. In this paper, the process of predicting water evaporation rate was formed into two types of problems: classification and prediction. The paper used multilayer perceptron neural network for the prediction purpose. Experiments show very promising and superior results.
In this paper, we proposed medical diagnostic system on mobile application using association rule extraction from text mining. The text mining was applied to achieve the knowledge discovery from historical medical data. This work aims to design and develop the system to medical diagnostic on mobile application supporting Thai language. Moreover, web application was built to manage knowledge achievement for using on mobile application by user or patient. The proposed system will discover the implication knowledge with association rules derived from Apiori algorithm to advice co-symptom of diagnosing to achieve more correctly on user's mobile. The implemented system was tested by user. The result shows that the proposed system can suggest co-symptom to achieve more correctness for medical diagnostic.
A new ATR tool for coastal surveillance radars is proposed. This tool is based on new features from target range, azimuth profiles and an adaptive neuro-fuzzy classifier with linguistic hedges (ANFC-LH). The tests with real data show that the ANFC-LH used in our study has a better classification result in comparison with several other ensemble classification algorithms such as bagged decision trees and adaptive boosting. Moreover, it is shown that the proposed ATR tool with ANFC-LH has a better performance than that one given in a recent publication.
This article reviews the characteristics of two common localization methods in ensemble Kalman filter (EnKF) systems: covariance localization (CL) and local analysis (LA). To obtain better assimilation results, a new data assimilation system coupled with fuzzy control algorithms is proposed, named CF (covariance fuzzy) and FA (fuzzy analysis). Motivated by fuzzy control concepts that have been developed in the control engineering field for years, the proposed methods improve the normal localization method which behaves like a Gaussian function but reaches zero at finite radius. To explore the effects of the two new algorithms on the background error covariance matrix and the gain matrix, numerical experiments are designed using a classical nonlinear model (i.e., the Lorenz-96 model). The experiments show that the new algorithms can eliminate spurious correlation of the background error covariance matrix. Additionally, with an increase of the assimilation intensity, the gain matrix followed by the update of the new algorithm. Finally, the experimental results demonstrate that the new algorithm has more robust performance than the common algorithm.
The inclusion of the chemistry field of study in higher education science and technology curricula aims to develop professionals who are able to analyze and solve multidisciplinary problems in a sustainable and correct way. Attending students to assess the role of chemistry in their education is critical to increasing success and improving their future professional practice. This article presents a Many-Valued Empirical Machine designed to capture Students' Perception of Chemistry in Higher Education Programs. The applied problem-solving method is based on a symbolic/sub-symbolic line of logical formalisms that articulate with an Artificial Neural Network approach to computing, being grounded on a view to knowledge representation and argumentation that considers not only the data entropic states but also its inherent Predicative Vagueness.
There are mainly two factors, multipath effect and strong background noise, affecting the performance of underwater weak acoustic signal detection. In this paper, to improve the performance, we propose a joint detection approach for underwater weak acoustic signal by combining the approaches of Passive Time Reversal (PTR) and Stochastic Resonance (SR). By calculating the input and output signal-to-noise ratios (SNR) theoretically, it's found that the proposed PTR-SR approach could improve the SNR of received signal, which is obtained by utilizing the multipath propagation channel and background noise simultaneously. Further, we propose a strategy to properly setting the free amplitude parameter Asr to optimize the SNR gain. Based on the Neyman-Pearson criterion, simulation results also highlight the performance of the proposed joint detection approach over single PTR approach and SR approach, especially in the circumstance of low SNR.
Elliptic curve cryptography is an important public-key cryptography following RSA. Here a new RNS (residue number system) approach for fast elliptic curve point multiplication over prime fields is proposed. The approach applies the Montgomery ladder for parallel elliptic curve point doublings (EC-doub) and point additions (EC-add). Meanwhile, residue number system with a wide dynamic range is used to support continuous multiplications, after which only one RNS Montgomery algorithm is enough to bring down the temporary results to valid range. In other words, both tasks can be implemented as many multiplications and a last reduction to improve performance. Also, Barrett modular reductions over small multipliers or modular multiplications over generalized Mersenne numbers can be applied for modular reductions over RNS moduli. A first analysis shows that the computation time for elliptic curve point multiplication over Fp can be reduced to a great extent at the cost of extra multipliers.
In recent years, deep learning has been applied to build soft sensor models. Compared to the classical latent variable models, the model based on deep learning has a good performance in tackling nonlinearity in process data. However, static soft sensor based on deep neural network (SSSDNN) fails to take into account the dynamic characteristics, which are unavoidable in some applications. To improve the performance of the soft sensor, a novel dynamic soft sensor model based on impulse response template and deep neural network (DSSDNN) is proposed by means of Wiener structure, and an iterative algorithm will be used to train this novel model. A case study based on real debutanizer column data demonstrates the desirable prediction performance of DSSDNN and shows that the DSSDNN is able to obtain better approximation accuracy than the static soft sensor model in nonlinear processes with dynamic characteristics.
This paper considers the feature selection scenario where features are generated sequentially and visible one by one and the full feature space is unknown in advance. Although there are some existing online streaming feature selection algorithms based on statistics or optimization, the main disadvantage they suffer is that a feature is permanently discarded once it is considered irrelevant even though it may become more relevant owing to the changing feature space. To overcome this drawback, our paper proposes a new online streaming feature selection algorithm based on a fixed-size buffer pool (BFS). Specifically, BFS maintains a buffer pool to dynamically retain and retrieve features to deal with the changing feature space and combines two different feature selector by a boost manner to improve the predictive performance. Extensive experiments are conducted on real-world datasets to evaluate the effectiveness of BFS and the results suggest that BFS is comparable to the state-of-the-art streaming feature selection algorithms, and requires less features or achieves higher predictive accuracy.
In this paper we introduce a stochastic time strength RBF neural network (ST-RBF) to prices forecasting of crude oil indices in which the network is set up by taking into consideration the impact of occurrence time for the historic data. A stochastic time-effective function is utilized to depict this, and a weight is assigned to each historic data, in which the time strength behavior is expressed through a combination of a Brownian function and a drift function. In empirical experiment, the data adopted is Chinese Daqing crude oil and Brent crude oil which is marked as one of the main global crude oil benchmarks. It is shown that the improved ST-RBF outperforms the traditional RBF neural network and increases the precision in predicting of crude oil prices.
Deep learning model has been extensively applied to the area of data classification and clustering in recently years because of its high similarity to human neurons in computational performance. In this paper, we propose a new deep model, multi-layer manifold based nonnegative matrix factorization with partial neural connections for image representation. In this model, a bidirectional multi-layer NMF decomposition for both basis and encoding vectors are conducted to capture structures in high dimensional data space. The connections of neurons in the learning are constrained in a neighborhood so that the similarities of elements in the same cluster can be learned. Test results on two different image datasets confirm that the proposed method can learn very good performance for image clustering tasks.