Search

Предметом вивчення в статті є інформаційна технологія обробки використання геопросторових інформаційних структур. Метою є удосконалення інформаційної технології обробки використання геопросторових інформаційних структур. Використовуваними методами є: методи теорії імовірності, математичної статистики, методи оптимізації, математичного моделювання та цифрової обробки зображень, аналітичні та емпіричні методи порівняльного дослідження. Отримані такі результати. В IDEF0-нотації описані етапи основного процесу прийняття управлінських рішень на основі аналізу геопросторової інформації. Наведена декомпозиція основного процесу прийняття управлінських рішень на основі аналізу геопросторової інформації, декомпозиція підпроцесу проведення тематичної інтерпретації видових зображень інформації, декомпозиція підпроцесу оцінювання апріорних імовірностей, декомпозиція підпроцесу прогнозування апріорних імовірностей та декомпозиція підпроцесу вибору оптимального рішення. Висновки. Удосконалено інформаційну технологію використання геопросторових інформаційних структур, яка, на відміну від відомих, базується на розроблених моделях формування видових зображень, тематичної інтерпретації видових зображень, прогнозування геопросторових даних та прийняття управлінських рішень і методах семантичної сегментації видових зображень та прийняття управлінських рішень на основі аналізу геопросторової інформації, що дозволяє формалізувати та автоматизувати процеси аналізу геопросторових даних, задовільнити зростаючі вимоги до систем обробки геопросторової інформації в умовах ризику та невизначеності та дозволяє приймати на їх основі обґрунтовані управлінські рішення.

The paper proposes an approach to development of a mathematical model for determination of monitoring objects using informativeness of their monitoring indicators. The possibility of calculating informativeness of an individual monitoring indicator has been investigated and lines of further research have been determined in terms of calculating the possibility of monitoring source (object) recognition error depending on relative informativeness of the signature and a priori probability of monitoring sources (objects).

The planning of drone trajectories in the search for ground-level objects is considered, in the case of challenging observation conditions. In existing planning algorithms, a priori data regarding the location of the objects are used in the secondary search, but no account is taken of their local visibility (on account of fog or smoke, say). To improve search productivity in challenging observation conditions, the observability may be taken into account in planning the drone trajectory. Heuristic models are used to assess the observability. The working trajectory is selected by calculating the maximum useful search information obtained from various possible trajectories. To eliminate Shannon indeterminacy in the utility of the information regarding successive points, we introduce an additional utility function. The results obtained by simulation of the search process confirm that this approach is more effective than trajectory planning on the basis of the maximum a priori probability that objects are present and on the basis of search entropy estimates.

In general, the industrial processes are semi-automatic, and are controlled by the operators. Since the operation principles of the industrial processes are complicated, it is difficult to label observations. The disturbances may be contained in the observations. Therefore, the unsupervised anomaly detection method is promising for research in the industrial processes. In the paper, a multivariate anomaly detection method is proposed, which is unsupervised and online. The priori probability of anomaly occurrence is necessary, and a hazard function selection method is defined at first. Secondly, Bayesian-based method is adopted for anomaly detection. In final, the Dempster-Shafer theory is introduced for fusing the univariate anomaly detection results. The numerical simulation is used for illustrating the anomaly detection power of the proposed method, and the TE process is implemented for testing the fault detection effectiveness. A real data set collected from a bathyscaphe is applied for demonstrating the power of leakage detection.

Missing data is a challenge in many applications, including intelligent transportation systems (ITS). In this paper, we study traffic speed and travel time estimations in ITS, where portions of the collected data are missing due to sensor instability and communication errors at collection points. These practical issues can be remediated by missing data analysis, which are mainly categorized as either statistical or machine learning(ML)-based approaches. Statistical methods require the prior probability distribution of the data which is unknown in our application. Therefore, we focus on an ML-based approach, Multi-Directional Recurrent Neural Network (M-RNN). M-RNN utilizes both temporal and spatial characteristics of the data. We evaluate the effectiveness of this approach on a TomTom dataset containing spatio-temporal measurements of average vehicle speed and travel time in the Greater Toronto Area (GTA). We evaluate the method under various conditions, where the results demonstrate that M-RNN outperforms existing solutions,e.g., spline interpolation and matrix completion, by up to 58% decreases in Root Mean Square Error (RMSE)., Comment: presented at the 2020 92nd IEEE conference on vehicular technology, 18 Nov.-16 Dec 2020 6 pages, 5 figures, 2 tables

Trip purpose imputation plays an important role in the development of smart passenger-centered services in Intelligent Transportation Systems. However, previous studies have paid relatively less attention to the trip purpose imputation at an individual level and required no real-time response. To narrow the gaps, a two-phase probabilistic framework TripImputor is proposed in this chapter, to perform real-time taxi trip purpose imputation and personalized recommendation service when passengers get off the taxi. To be more specific, a two-stage clustering algorithm is proposed in the first phase to obtain candidate activity areas (CAAs) in the urban space. Based on that, we extract fine-grained spatiotemporal regularity of human behaviors within each CAA with the Foursquare check-in data. Such process can gain on the priori probability for each kind of human activity and obtain the posterior probabilities of taxi trip purposes by employing Bayes’ theory. In the second phase, historical drop-off locations would be clustered then matched with CAAs for the real-time response. Finally, extensive experiments are conducted to evaluate the effectiveness and response of our proposed methods. The results show that the proposed methods can present accurate trip purposes and offer the recommendation information within 1.6 s averagely.

Travel behavior understanding is a long-standing and critically important topic in the area of smart cities. Big volumes of various GPS-based travel data can be easily collected, among which the taxi GPS trajectory data is a typical example. However, in GPS trajectory data, there is usually little information on travelers’ activities, thereby they can only support limited applications. Quite a few studies have been focused on enriching the semantic meaning for raw data, such as travel mode/purpose inferring. Unfortunately, trip purpose imputation receives relatively less attention and requires no real-time response. To narrow the gap, we propose a probabilistic two-phase framework named TripImputor , for making the real-time taxi trip purpose imputation and recommending services to passengers at their dropoff points. Specifically, in the first phase, we propose a two-stage clustering algorithm to identify candidate activity areas (CAAs) in the urban space. Then, we extract fine-granularity spatial and temporal patterns of human behaviors inside the CAAs from foursquare check-in data to approximate the priori probability for each activity, and compute the posterior probabilities (i.e., infer the trip purposes) using Bayes’ theorem. In the second phase, we take a sophisticated procedure that clusters historical dropoff points and matches the dropoff clusters and CAAs to immerse the real-time response. Finally, we evaluate the effectiveness and efficiency of the proposed two-phase framework using real-world data sets, which consist of road network, check-in data generated by over 38 000 users in one year, and the large-scale taxi trip data generated by over 19 000 taxis in a month in Manhattan, New York City, USA. Experimental results demonstrate that the system is able to infer the trip purpose accurately, and can provide recommendation results to passengers within 1.6 s in Manhattan on average, just using a single normal PC.

The distributed intelligent pension system is a new old-age pension system that is designed to solve the problem existed in decentralized management system in traditional nursing homes, such as information isolation and imperfect pension facilities. The system combines the advantages of RFID technology and video linkage monitoring. In order to know whether the elderly is well taken care of, the two types of information need to be processed and analyzed. Data fusion technology is an effective tool to solve the optimal decision of multi attribute data. In the algorithm of data fusion, the neural network algorithm has good fault tolerance and adaptability, and requires a small priori probability distribution of the system. It can handle incomplete and inaccurate information. Combined with the multi-source and massive characteristics of the data of distributed intelligent pension system, the data processing has the characteristics of real-time and accuracy. In addition, the BP neural network has the characteristics of simple realization and high recognition precision in a certain range. The BP neural network algorithm is used as the research, and the additional momentum method is used to improve the traditional BP algorithm. In the same direction, the gradient is added to the weight and threshold, and the algorithm is guaranteed to the direction of convergence.

Accuracy has been used traditionally to evaluate the performance of classifiers. However, it is well known that accuracy is not able to capture all the different factors that characterize the performance of a multiclass classifier. In this manuscript, accuracy is studied and analyzed as a weighted average of the classification rate of each class. This perspective allows us to propose the dispersion of the classification rate of each class as its complementary measure. In this sense, a graphical performance metric, which is defined in a two dimensional space composed by accuracy and dispersion, is proposed to evaluate the performance of classifiers. We show that the combined values of accuracy and dispersion must fall within a clearly bounded two dimensional region, different for each problem. The nature of this region depends only on the a priori probability of each class, and not on the classifier used. Thus, the performance of multiclassifiers is represented in a two dimensional space where the models can be compared in a more fair manner, providing greater awareness of the strategies that are more accurate when trying to improve the performance of a classifier. Furthermore we experimentally analyze the behavior of seven different performance metrics based on the computation of the confusion matrix values in several scenarios, identifying clusters and relationships between measures. As shown in the experimentation, the graphical metric proposed is specially suitable in challenging, highly imbalanced and with a high number of classes datasets. The approach proposed is a novel point of view to address the evaluation of multiclassifiers and it is an alternative to other evaluation measures used in machine learning.

Vehicle shape information is very important in Intelligent Traffic Systems (ITS). In this paper, we present a way to exploit a training data set of vehicles released in different years and captured under different perspectives. Also the efficacy of clustering to enhance the make/model classification is presented. Both steps led to improved classification results and a greater robustness. Deeper convolutional neural network based on ResNet architecture has been designed for the training of the vehicle make/model classification. The unequal class distribution of training data produces an a priori probability. Its elimination, obtained by removing of the bias and through hard normalization of the centroids in the classification layer, improves the classification results. A developed application has been used to test the vehicle re-identification on video data manually based on make/model and color classification. This work was partially funded under the grant.

We tackle the complex problem of determining entailment relationships between case law documents, one of the tasks in the Competition on Legal Information Extraction and Entailment (COLIEE). With input of an entailed fragment from a case coupled with a candidate entailing paragraph from a noticed case, our approach relies on four main components: (1) extraction of similarity measures between the two pieces of text; (2) application of a transformer-based technique on the input text; (3) applying a threshold-based classifier; and (4) post-processing the results considering the a priori probability determined by the data distribution on the training samples and combining the results of (1) and (2). Our experiments achieved an F-score of 0.70 on the official COLIEE test dataset, ranking first among all competitors for that task in the 2019 competition.

Multi-sensor data fusion (MSDF) is an efficient technology to enhance the performance of the system with the involvement of different kinds of sensors, which are broadly utilized in many fields at present. However, the data obtained from multi-sensors may have different degrees of uncertainty in the practical applications. Evidence theory is very useful to convey and manage uncertainty without a priori probability, so that it has been proverbially adopted in the information fusion fields. However, in the face of conflicting evidences, it has the possibility of producing counterintuitive results via conducting the Dempster's combination rule (DCR). To solve the above-mentioned issue, a hybrid MSDF method is exploited through integrating a newly defined evidential credibility measure of evidences based on prospect theory and the evidence theory. More specifically, a series of concepts for the evidential credibility measure are first presented, including the local credibility degree, global credibility degree, evidential credibility estimation and credibility prospect value function to comprehensively describe the award and punish grades in terms of credible evidence and incredible evidence, respectively. Based on the above researches, an appropriate weight for each evidence can be obtained. Ultimately, the weight of each evidence is leveraged to amend the primitive evidences before conducting DCR. The results attained in the experiments demonstrate that the hybrid MSDF approach is efficient and superior to handle conflict evidences as well as the application in data fusion problems.

In this work, we address the problem of reducing the false positives for object detection in videos. We employ the motion cue to build a foreground probability model. Then the mean expectation of the pixel-level foreground probability is computed to assign a priori probability to the sliding window in detection. The proposed foreground model is evaluated with the detection framework of Deformable Part Models (DPM). We combine the response of DPM detector and the mean probability expectation to form the features and train a linear classifier. The proposed approach is threshold-free, and reduces the false positives in object detection by the foreground cues. Besides, we describe an integral probability image for fast computation of the mean probability expectation. Experimental results show that the proposed method achieve superior performance over the baseline of Deformable Part Models.

Based on Bayesian classification algorithm principle and implementation, propose an improved method of the algorithm. Firstly, instead of constant probability of spam, actual priori probability is used. Secondly, the selective range and rule of token is improved. Finally, add URLs and images into detection content. A mail recognizer based on improved Bayesian classification is designed. The experiment result shows that the improved Bayesian classification algorithm works well in practice.

Transfer learning is a method that studies how to identify the useful knowledge and skills in the previous tasks, and uses them to the new tasks or domains. At present, the research on transfer learning mostly focuses on the field of long texts. However, the source data should be given for the transportation from long texts to the sh ort ones, and the priori probability distribution of the data should be given at the same time. In order to solve the problems, the algorithm which is called FSFP (Free Source selection Free Priori probability distribution) is proposed. It can transfer knowledge from the long texts to the short ones. Latent semantic analysis is used to extract the key words as seed characteristic sets, which are semantically related to the long texts from the target domain. And then the graph structure of online information is built. With the help of the improved Laplacian Eigenmaps, the feature representations of high- dimensional data are mapped to a low-dimensional space. Lastly, the target data are classified in the constraint of minimizing the mutual information between the instance and the feature representation. The experimental results on large data sets show the effectiveness of the new algorithm.

Part 4: Privacy II; International audience; This paper proposes a new privacy-preserving scheme for estimating the size of the intersection of two given secret subsets. Given the inner product of two Bloom filters (BFs) of the given sets, the proposed scheme applies Bayesian estimation under assumption of beta distribution for an a priori probability of the size to be estimated. The BF retains the communication complexity and the Bayesian estimation improves the estimation accuracy.An possible application of the proposed protocol is an epidemiological datasets regarding two attributes, Helicobactor pylori infection and stomach cancer. Assuming information related to Helicobactor Pylori infection and stomach cancer are separately collected, the protocol demonstrates that a χ2-test can be performed without disclosing the contents of the two confidential databases.

We argue that in spite of their apparent dissimilarity, the methodologies employed in the a priori and a posteriori assessment of probabilities can both be justified by appeal to a single principle of inductive reasoning, viz., the principle of symmetry. The difference between these two methodologies consists in the way in which information about the single-trial probabilities in a repeatable chance process is extracted from the constraints imposed by this principle. In the case of a posteriori reasoning, these constraints inform the analysis by fixing an a posteriori determinant of the probabilities, whereas, in the case of a priori reasoning, they imply certain claims which then serve as the basis for subsequent probabilistic deductions. In a given context of inquiry, the particular form which a priori or a posteriori reason may take depends, in large part, on the strength of the underlying symmetry assumed: the stronger the symmetry, the more information can be acquired a priori and the less information about the long-run behavior of the process is needed for an a posteriori assessment of the probabilities. In the context of this framework, frequency-based reasoning emerges as a limiting case of a posteriori reasoning, and reasoning about simple games of chance, as a limiting case of a priori reasoning. Between these two extremes, both a priori and a posteriori reasoning can take a variety of intermediate forms.

Bayesian filtering approach is widely used in the field of anti-spam now. However, the two assumptions of this algorithm are significantly different with the actual situation so as to reduce the accuracy of the algorithm. This paper proposes a detailed improvement on researching of Bayesian Filtering Algorithm principle and implement method. It changes the priori probability of spam from constant figure to the actual probability, improves selection and selection rules of the token, and also adds URL and pictures to the detection content. Finally it designs a spam filter based on improved Bayesian filter approach. The experimental result of this improved Bayesian Filter approach indicates that it has a beneficial effect in the spam filter application.

Nowadays imbalanced learning represents one of the most vividly discussed challenges in machine learning. In these scenarios, one or some of the classes in the problem have a significantly lower a priori probability, usually leading to trivial or non-desirable classifiers. Because of this, imbalanced learning has been researched to a great extent by means of different approaches. Recently, the focus has switched from binary classification to other paradigms where imbalanced data also arise, such as ordinal classification. This paper tests the application of learning pairwise ranking with multiple granularity levels in an ordinal and imbalanced classification problem where the aim is to construct an accurate model for donor-recipient allocation in liver transplantation. Our experiments show that approaching the problem as ranking solves the imbalance issue and leads to a competitive performance.

When the influencing factors are not complete, the method based on mathematical statistics and probability theory is difficult to calculate the uncertainty of the competition results. So the calculation method of uncertainty for competition prediction based on the evidence theory features and application scenes is used. In this paper, firstly we use the improved D-S evidence theory to calculate the uncertainty of the data and the unknown a priori probability. Secondly, the paper further presents a calculation result of the tennis match, in which the condition is that the influence factors produce the contradiction or the initialization input is not sure. Experimental results show that the method proposed in this paper makes up for the lack of initial conditions and leads to the uncertainty of the competitive results. On the basis of the previous method, it can not be calculated in the absence of a data source. It is more realistic to consider the influence factors in the tennis movement, and it has better applicability.

Email is an incontestable communication mode in both professional and personal correspondences. The survey shows that an ordinary white-collar worker spend at least an hour every day to deal with the email. Handling spam which is disguised as normal email is waste our time. In this paper, we propose a spam detection method upon statistical-based Bayesian algorithm. Firstly, the method use actual priori probability of spam instead of constant probability. Secondly, the selective range and rules of tokens is improved. Finally, our method add URLs and images into detection content. The experiment result shows that the improved statistical-based Bayesian classification algorithm works well in practice.

Optimal team decision making subject to error-prone team members with different capabilities has been studied extensively - particularly in the context of binary classification. The over-arching goal is to correctly classify an object as either being a True Target or a False Target. Each team member comes with pre-specified Type I and II error rates and is asked whether or not he determines the object to be a True Target. Based on the member responses, a collective decision that minimizes the probability of misclassification (or some other suitable metric) is made. This is a widely researched topic, having applications in pattern recognition, organizational decision making, social (dichotomous) choice situations, reliability studies etc.; however, in past work, the obvious connection to information theory is missing. In this work, we establish the optimal team decision rules that minimizes the probability of misclassification, by direct application of Bayesian decision theory. In doing so, we bring out the key role played by the a priori probability (prior) that the object is a True Target. In particular, for a homogeneous team composition, we establish the criteria whence the majority voting scheme is indeed optimal. Whereupon, it immediately follows that the higher the prior, the fewer the number of affirmative votes needed to classify the object as a True Target.

Bayesian formula is used to determine diagnosis sequence when several fault trees meet requirements. Bayesian prior probability is usually determined through expert or the user's subjective judgment and historical experience. If there is lack of expert experience, the determination of priori probability is very difficult. A real-time priori probability calculation method is proposed, which needn’t any priori-knowledge and can regulate automatic on the monitoring parameters. It takes into account the multiple diagnosis impact and more flexible than fixed priori probability according application.

In the information retrieval framework, there are problems where the goal is to recover objects of a particular class from big sets of un labelled objects. In some of these problems, only examples from the class we want to recover are available. For such problems, the machine learning community has developed algorithms that are able to learn binary classifiers in the absence of negative examples. Among them, we can find the positive Bayesian network classifiers, algorithms that induce Bayesian network classifiers from positive and un labelled examples. The main drawback of these algorithms is that they require some previous knowledge about the a priori probability distribution of the class. In this paper, we propose a wrapper approach to tackle the learning when no such information is available, setting this probability at the optimal value in terms of the recovery of positive examples. The evaluation of classifiers in positive un labelled learning problems is a non-trivial question. We have also worked on this problem, and we have proposed a new guiding metric to be used in the search for the optimal a priori probability of the positive class that we have called the pseudo F. We have empirically tested the proposed metric and the wrapper classifiers on both synthetic and real-life datasets. The results obtained in this empirical comparison show that the wrapper Bayesian network classifiers provide competitive results, particularly when the actual a priori probability of the positive class is high.

A probabilistic classification framework is presented that can combine temporal contextual information derived from an existing land-cover map in order to improve the classification accuracy of land-cover classes that can not be discriminated well when using spectral information only. The transition probability is computed by using the existing land-cover map and training data, and considered as a priori probability. By combining the a priori probability with conditional probability computed from spectral information via a Bayesian combination rule, the a posteriori probability is finally computed and then the final land-cover types are determined. The method presented in this paper can be adopted to any probabilistic classification algorithms in a simple way, compared with conventional classification methods that require heavy computational loads to incorporate the temporal contextual information. A case study for crop classification using time-series MODIS data sets is carried out to illustrate the applicability of the presented method. The classification accuracies of the land-cover classes, which showed lower classification accuracies when using only spectral information due to the low resolution MODIS data, were much improved by combining the temporal contextual information. It is expected that the presented probabilistic method would be useful both for updating the existing past land-cover maps, and for improving the classification accuracy.

The distinction between a priori and a posteriori knowledge has been the subject of an enormous amount of discussion, but the literature is biased against recognizing the intimate relationship between these forms of knowledge. For instance, it seems to be almost impossible to find a sample of pure a priori or a posteriori knowledge. In this paper it will be suggested that distinguishing between a priori and a posteriori is more problematic than is often suggested, and that a priori and a posteriori resources are in fact used in parallel. We will define this relationship between a priori and a posteriori knowledge as the bootstrapping relationship. As we will see, this relationship gives us reasons to seek for an altogether novel definition of a priori and a posteriori knowledge. Specifically, we will have to analyse the relationship between a priori knowledge and a priori reasoning, and it will be suggested that the latter serves as a more promising starting point for the analysis of aprioricity. We will also analyse a number of examples from the natural sciences and consider the role of a priori reasoning in these examples. The focus of this paper is the analysis of the concepts of a priori and a posteriori knowledge rather than the epistemic domain of a posteriori and a priori justification.

In order to identify the threat of computer network security and evaluate its fragility comprehensively, the related factors of network security are studied, and the methods based on attack graph are improved. Based on the attribute attack graph, the probabilistic attack graph model is generated by adding various factors which affect network security. The model uses security equipment performance data, common vulnerability scoring system data and etc. to calculate priori probability, finally obtains the network security index, and carries on the exploratory analysis. The experimental results show that the model is feasible and effective. Compared with other vulnerability assessment methods, the model has the characteristics of comprehensive evaluation and concise calculation.

A binary classification method, Probabilistic Extreme Learning Machine (called P-ELM), is proposed to enhance the reliability of the classification of an unknown object. P-ELM method integrates ELM, density methods and Bayesian decision theory in order to take into account a priori probability of the process and the uncertainty of the ELM predictions. The P-ELM algorithm may inhibit uncertainty of the extreme learning machine prediction in the different trials of simulation due to the initialization of input weights and bias, which would damage the reliability of the classification for the new objects. Simulations results from a municipal wastewater treatment plant show that the proposed P-ELM binary classification method can provide the reliability of the classification for those samples near the boundaries of the classes and the reliability and accuracy outperform the ELM model.

Motivation: With the wealth of sequence data and the huge amount of data generated from molecular technologies, the issue of gene classification/prediction has become a central challenge in the field of microarray data analysis. This has led to the application of many well-established supervised learning (SL) algorithms in an attempt to provide more accurate and automatic diagnosis class (cancer/non cancer) prediction. Virtually all research on SL addresses the task of learning to classify complete domain instances. However, in some research situations we often have to classify instances given incomplete vectors, which can affect the predictive accuracy of learned classifiers. The task of learning an accurate incomplete data classifier from instances raises a number of new issues some of which have not been properly addressed by bioinformatics research. Thus, an effective missing value estimation method is required for improving predictive accuracy. Results: The essence of the approach is the proposal that prediction using supervised learning can be improved in probabilistic terms given incomplete microarray data. This imputation approach is based on the a priori probability of each value determined from the instances at that node of a decision tree (PDT) that have specified values. The proposed approach exploits the total probability and Bayes' theorems and it has three versions. We evaluate our approach with other supervised learning techniques including C5.0, classification and regression trees (CART), k-nearest neighbour (k-NN), linear discrimination (LD) naive Bayes classifier (NBC), Repeated Incremental Pruning to Produce Error Reduction (RIPPER) and support vector machines (SVMs), from the point of view of their effect or tolerance of incomplete test data. Eight cancer related gene expression datasets are utilized for this task. Experimental results are provided to illustrate the efficiency and the robustness of the proposed algorithm.

Bayes's theorem is predominantly used in diagnosing based on the results of various diagnostic tests. This statistical approach is intuitive in differential diagnosis as it explicitly takes into consideration data from medical history, physical examination, laboratory findings and imaging. Bayes's theorem states that the probability of disease occurrence (or occurrence of other outcome) after new information is obtained, called a posteriori probability, depends directly on an a priori probability and the value of likelihood ratio associated with a given test result. This paper describes basic Bayesian analysis in relation to the diagnosis of two types of secondary hypertension; primary aldosteronism and pheochromocytoma. This choice is based on two facts; primary aldosteronism is believed to be the most common and the most commonly detected cause of symptomatic hypertension and pheochromocytoma is thought to have rapid progress and stormy clinical course. This article aims to draw physicians' attention to and increase the knowledge of Bayesian analysis, and to describe its use in everyday clinical decision making. On the basis of this theorem's foundations, the discussion in relation to the issue of differential diagnosis between physicians, their patients, and medical students should also improve. When used in practice, one should be aware, however, of Bayesian analysis limitations concerning the diagnostic test application and limited knowledge of diagnostic test accuracy, and insecure or faulty a priori probability estimates.

An adaptive classifier designed for high-dimensional image data classification has recently been proposed. The main feature in this proposed classifier consists of a novel method to acquire semilabeled samples in an iterative way during the classification process itself. The semilabeled samples are then added to the set of available labeled samples, thus increasing the total number of training samples and consequently the reliability of the estimated values for the parameters in a parametric classifier. In this study we propose the insertion of an additional step into the iterative procedure implemented in the adaptive classifier to produce more realistic estimates of the a priori probability of the classes. The proposed procedure to estimate the a priori probability of the classes makes use of the same weighted approach used by the adaptive classifier, is computationally efficient, and can be easily implemented. Tests using hyperspectral image data have shown that the proposed approach can contribute to ...

Spam is one of the major problem of today's life because it causes a lot of extra expense both in network infrastructure and our individual life. Among those approaches developed to detect spam, the content-based detection technique, especially statistical-based Bayesian algorithm is important and popular. However, the basic Bayesian algorithm permits on assumption and estimation. In this paper, we proposed an improved method to increase the accuracy of the algorithm. Firstly, use actual priori probability instead of constant probability of spam. Secondly, expand the selective range and rule of tokens. Finally, add URLs and images into detection content. Keywords-spam; statistical-based bayesian classification; content detection

Numerous classification algorithms have been developed, many of which are highly specific and only solve a reduced class of problems. Maximum likelihood classification (MLC) is the most widely used classification method. The underlying assumption in performing MLC is that the prior probability of land cover is equal. However, a priori occurrence probability has a crucial effect on classification results. The objective of this paper is to improve the accuracy of MLC using a priori information in a knowledge-based system. The mathematical formulations and the strategy are presented. Estimates of a prior probability through crop areas, crop calendar, and some a priori probability about agricultural practices have been used in augmenting the probability of pixels. An industrial agricultural field, Moghan Plain in northwestern Iran, has been selected for testing. A total of 176 ground truth points were identified and measured in the field, and 323 fields were used for accuracy evaluation. Prior probabilities w...

The positive unlabeled learning term refers to the binary classification problem in the absence of negative examples. When only positive and unlabeled instances are available, semi-supervised classification algorithms cannot be directly applied, and thus new algorithms are required. One of these positive unlabeled learning algorithms is the positive naive Bayes (PNB), which is an adaptation of the naïve Bayes induction algorithm that does not require negative instances. In this work we propose two ways of enhancing this algorithm. On one hand, we have taken the concept behind PNB one step further, proposing a procedure to build more complex Bayesian classifiers in the absence of negative instances. We present a new algorithm (named positive tree augmented naive Bayes, PTAN) to obtain tree augmented naive Bayes models in the positive unlabeled domain. On the other hand, we propose a new Bayesian approach to deal with the a priori probability of the positive class that models the uncertainty over this parameter by means of a Beta distribution. This approach is applied to both PNB and PTAN, resulting in two new algorithms. The four algorithms are empirically compared in positive unlabelled learning problems based on real and synthetic databases. The results obtained in these comparisons suggest that, when the predicting variables are not conditionally independent given the class, the extension of PNB to more complex networks increases the classification performance. They also show that our Bayesian approach to the a priori probability of the positive class can improve the results obtained by PNB and PTAN.

In Bayesian probabilistic approach for uncertain reasoning, one basic assumption is that a priori knowledge about the uncertain variable is modeled by a probability distribution. When new evidence representable by a constant set is available, the Bayesian conditioning is used to update a priori knowledge. In the conventional D-S evidence theory, all bodies of evidence about the uncertain variable are imprecise and uncertain. All bodies of evidence are combined by so-called Dempster's rule of combination to achieve a combined body of evidence without considering a priori knowledge. From our point of view, when identifying the true value of an uncertain variable, Bayesian approach and evidence theory can cooperate to deal with uncertain reasoning. Firstly all imprecise and uncertain bodies of evidence about the uncertain variable are fused to achieve a combined evidence based on a priori knowledge, then the a posteriori probability distribution is achieved from a priori probability distribution by conditioning on the combined evidence. In this paper we firstly deal with the knowledge updating problem where a priori knowledge is represented by a probability distribution and new evidence is represented by a random set. Then we review the conditional evidence theory which resolves the knowledge combining problem based on a priori probabilistic knowledge. Finally we discuss the close relationship between knowledge updating procedure and knowledge combining procedure presented in this paper. We show that a posteriori probability conditioned on fused body of evidence satisfies the Bayesian parallel combination rule.

Based on the hypothesis that the sound of the infant cry contains information on the infant's health status, research is done to improve classi¯cation of neonate crying sounds into categories called 'normal' and 'abnormal' - the latter referring to some hypoxia- related disorder. Research in this field is hindered by lack of test cases and limited understanding of feature relevance. The research described here combines various ways of dealing with the small data set problem. First, feature pre-selection is done using sequential backwards elimination of possible combinations where the performance of the set of features is tested by a Probabilistic Neural Network which has the advantage of fast learning. These features are then fed into a neural network committee consisting of Radial Basis Function Neural Networks. Bootstrapping is used to generate the committee. This construction yields a multi-classifier system with an overall classi¯cation performance of 85% on the data set, an increase of 34% with respect to the a priori probability of 51%. Several leave-1-out experiments for Linear Discriminant Analysis (LDA), Support Vector Machines (SVM) and Neural Networks (NN) have been conducted in order to compare the performance of the multi-classifier system. Keywords: Infant Cry Analysis; Feature reduction, Neural Networks; Support Vector Machines

The Information Fusion (IF) process is becoming increasingly more sophisticated, particularly through the incorporation of methods for high-level reasoning when applied to the situation analysis domain. A fundamental component of the IF process is a database (or databases) containing a priori knowledge that lists expected objects, behaviors of objects, and relationships between objects as well as all the possible attributes that can be inferred from measurements coming from a given sensor suite. We first present the basic concept of an existing support database (consisting of more than 2200 platforms) for Identity information fusion, and discuss its extension for higher-level fusion (e.g. situation and threat assessment). The database contains all the salient features needed for refining the identity of any target by the fusion of sensor information, and for addressing the situation and threat posed by groups of objects. The database is especially well suited for use in a Dempster-Shafer evidential reasoning scheme although it can also be used with Bayesian reasoning, if a priori probability distributions are known. Convincing results on several realistic scenarios of Maritime Air Area Operations and Direct Fleet Support are presented. This paper then develops the advanced concept of a Knowledge Management and Exploitation Server (KNOWMES) to support the IF process, through the use of ontologies and heterogeneous knowledge sources, which are necessary for higher level fusion.

Molecular bioinformatics is a transdisciplinary working area. One hot topic is the design of drugs using computers and intelligent algorithms. This is known as in silico approach. We use a new in silico approach for separating active ligand molecules from inactive ones for different drug targets. This kind of retrospective virtual screening is performed by using encoded molecule data and a neuro-fuzzy methodology for classification, feature selection, and rule generation. We generate rules in a retrospective screening process that identify regions, where clearly more active compounds can be found compared to their a priori probability. We show that our approach is superior to a common descriptor-based standard technique.

Summary We address the question of introducing expert knowledge in a hierarchical Bayesian spatiotemporal model. Specifically, we combine avalanche count data and an expert prior judgement about avalanche climatology to cluster French Alpine townships into two coherent groups. A spatial regression defines the a priori probability of belonging to each group and an innovative method is proposed to elicit its parameters by working with an expert on replicates of simulated maps. As a benefit of the extra data knowledge, we show the existence of robust north–south clusters of decreasing–increasing avalanche activity resulting from the interaction of local climate change patterns with altitude distribution. Spatial dependence of avalanche activity has also been inferred and is characterized by an anisotropy, with longer dependence along the north-west–south-east axis.

The problem of the congruence of the design–product system is studied in this work, as a particular case of more complex designing–production integration. The proposed approach is based on entropy principles. An a priori probability distribution regarding design and an a posteriori probability distribution regarding the product are assumed. The distance between such distributions permits us to evaluate the above congruence. The proposed model, which is made more user friendly by transforming the probability approach into a fuzzy approach, is a useful tool for the control of production systems, e.g. for the variation control of features.

This paper proposes a solution which is called Analysis of Key Events in Audit Data Using Dynamic-Created Influence Nets. According to the audit transaction and the events operating results the prior probability of actionable events and baseline probability of desired events are obtained statistically, the causal strength between audit events and the corresponding event type is also obtained. Then the influence net is created to analyze sensitivity of the audit events in their types and search set of key events. We apply the proposed method in audit data of a management system. The experimental results show that the method is useful to get key events in complex audit data.

Given a labeled Petri net with silent (unobservable) transitions, we are interested in performing initial marking estimation in a probabilistic setting. We assume a known finite set of initial markings, each with some a priori probability, and our goal is to obtain the conditional probabilities of initial markings of the Petri net, conditioned on an observed sequence of labels. Under a Markovian assumption on the probabilistic model, we develop a recursive algorithm that allows us to efficiently determine the conditional probabilities for each possible initial marking (conditioned on the sequence of observations seen so far). We illustrate the proposed methodology via an example and discuss potential applications in the context of initial state opacity for security applications. © 2014 IEEE. 2015-February 6725 6730

A transient power quality assessment method is proposed in this paper, using Naive B ayes classification method which is based on big data processing architecture. In this architecture, data sources will be extended to the aspects of power grid monitoring data, the power customer data and the public data, and the assessment severity will be classified into the normal state, the abnormal state, the critical state, and the failed state, according to the Naive Bayes classification results. Based on the data type of transient power quality assessment, big data processing architecture used in this paper can be able to process distributed data and streaming data, so that it can ensure not only updates classifier rules regularly, but also the real-time condition assessment. In the classifier training phase, we use the massive historical data as the distributed learning object, and generate assessment rules periodically. In the state assessment phase, each assessment node will update the assessment rules generated by training phase, generate real­ time evaluation of samples from stream processing fram ework, and evaluate the power quality state according to the current rule. On this basis, this paper designs a Naive Bayes classification method based on MapReduce processing, and realizes the map and reduce process method to compute the priori probability and the conditional probability in distributed way. Experiments show that the transient power quality evaluation method based on the big data analysis presented in this paper is feasible, and achieve good results both in classification accuracy and processing speed.

International audience; In this paper we propose an approach that exploits social data associated with a Web resource to measure its a priori relevance. We show how these interaction traces left by the users on the resources, which are in the form of social signals as the number of like and share, can be exploited to quantify social properties such as popularity and reputation. We propose to model these properties as a priori probability that we integrate into language model. We evaluated the effectiveness of our approach on IMDb dataset containing 167438 resources and their social signals collected from several social networks. Our experimental results are statistically significant and show the interest of integrating social properties in a search model to enhance the information retrieval.

Unreplicated factorial designs are widely used for designed experimentation in industry. In the analysis of designed experiments, the experimental factors influencing the response must be identified and separated from those that do not. An abundance of procedures intended to perform this selection have been introduced in the literature. A recent study indicated that the procedure due to Box and Meyer outperforms the lot of the other selection procedures in terms of efficiency and robustness. The procedure of Box and Meyer rests on a quasi-Bayesian foundation and utilizes generic domain knowledge, in the form of a common-for-all-factors a priori probability, that a factor significantly influences the response, to calculate an a posteriori probability for each factor. This paper suggests a strategy for introducing more elaborate domain knowledge about the experimental factors in the procedure of Box and Meyer, aiming to further improve its performance.

The traditional equating of the a priori and the necessary was challenged by Kripke, who indicated the possibility of a priori knowledge of contingently true sentences, as well as a posteriori knowledge of necessarily true sentences. This note discusses a new species of sentences with such properties. One example of this is (1) This sentence is necessarily false which appears to be a contingently false sentence, known a priori, although it is hard to see when it might be true. Further examples are considered, and it is argued that the notions of being known a priori and aposteriori should be examined further, along with the notions of necessary and contingent truths.

Co-activation of distinct brain regions is a measure of functional interaction, or connectivity, between those regions. The co-activation pattern of a given region can be investigated using seed-based activation likelihood estimation meta-analysis of functional neuroimaging data stored in databases such as BrainMap. This method reveals inter-regional functional connectivity by determining brain regions that are consistently co-activated with a given region of interest (the “seed”) across a broad range of experiments. In current implementations of this meta-analytic connectivity modeling (MACM), significant spatial convergence (i.e. consistent co-activation) is distinguished from noise by comparing it against an unbiased null-distribution of random spatial associations between experiments according to which all gray-matter voxels have the same chance of convergence. As the a priori probability of finding activation in different voxels markedly differs across the brain, computing such a quasi-rectangular null-distribution renders the detection of significant convergence more likely in those voxels that are frequently activated. Here, we propose and test a modified MACM approach that takes this activation frequency bias into account. In this new specific co-activation likelihood estimation (SCALE) algorithm, a null-distribution is generated that reflects the base rate of reporting activation in any given voxel and thus equalizes the a priori chance of finding across-study convergence in each voxel of the brain. Using four exemplary seed regions (right visual area V4, left anterior insula, right intraparietal sulcus, and subgenual cingulum), our tests corroborated the enhanced specificity of the modified algorithm, indicating that SCALE may be especially useful for delineating distinct core networks of co-activation.

The Context-Dependent Deep-Neural-Network HMM, or CD-DNN-HMM, is a powerful acoustic modeling technique. Its training process typically involves unsupervised pre-training and supervised fine-tuning. In the paper, we demonstrate that the performance of DNNs can be improved by utilizing a large amount of unlabeled data in the training procedure. In our method, CD-DNN-HMM trained using 309 hours of unlabeled data and 24 hours of labeled data achieved word-error rate of 23.7% on the Hub5'00-SWB phone-call transcription task, compared to word-error rate of 24.3% obtained by a CD-DNN-HMM trained without using unlabeled data. We also applied a priori probability smoothing algorithm that further reduced the error rate to 23.2%. On RT03S-FSH benchmark corpus, our experimental results show that similar performance gains can be obtained by the use of unlabeled data.

This paper describes conditioned Dempster-Shafer (CDS) theory, a probabilistic calculus for dealing with possibly non-Bayesian evidence when the underlying a priori knowledge base is possibly non-Bayesian. Specifically, we show that the Dempster-Shafer combination operator can be "conditioned" to reflect the influence of any a priori knowledge base which can be modeled by a Dempster-Shafer belief measure. We show that CDS is firmly grounded in probability theory-specifically, in the theory of random sets. We also show that it is a generalization of the Bayesian theory to the case when both evidence and a priori knowledge are ambiguous. We derive the algebraic properties of the theory when a priori knowledge is assumed fixed. Under this assumption, we also derive the form of CDS in the special case when fixed a priori knowledge is Bayesian.

The authors think that the main problem with the application of Bayes' theorem in medical expert systems is that the formula itself cannot resolve the contradiction between frequency of disease manifestation and its specificity, especially for those manifestations that have a very low frequency but a high specificity. The authors have found in practice that the weighted summation method in its various application forms can handle these situations well. The diagnostic weight value of any manifestation can be adjusted by clinical experts, based on its actual importance for the diagnosis, as judged by theory and practical experiences. This method will be discussed more in detail in the authors' other paper. When adopting Bayes' method as the algorithm of a medical expert system, one must ensure that the theorem's two assumptions are satisfied as best as possible. There should be a sufficient quantity of cases of the diseases in the database, and the actual information should be as accurate as possible. When a system based on Bayes' formula is used in epidemiologically different populations, the a priori probability of the corresponding disease should be corrected. Especially, those manifestations with high specificity should be chosen as diagnostic indicators. In summary, applying Bayes' method can provide quite satisfactory results in many applications.