This paper introduces Norma, a novel association analysis framework for continuous spatial variables. Unlike other point-based spatial analysis methods, Norma associates a point-wise function with each continuous variable-e.g., a function that returns poverty rates for each location in New Mexico, and then finds interesting associations between continuous variables by analyzing relationships between the respective point-wise functions. Moreover, the paper introduces a new association called Continuous Variable Threshold (CVT) pattern, aiming to identify a pair of thresholds within the domains of two continuous variables which exhibit strong associations within an observation area. For example, it may unveil a strong association between COVID-19 infection rates above 2% and poverty rates above 15% in New Mexico. To find such associations, Norma employs a novel interestingness function, which measures agreement with respect to hotspots where each point-wise function exceeds the associated threshold. To be able to compute such hotspots, the paper proposes a novel grid-based spatial hotspot-growing algorithm which computes regions of a point-wise function above a given threshold. Furthermore, Norma introduces a measure area under the curve (AUC) for assessing variable relatedness based on observed CVT associations and an algorithm to determine the AUC in 2D space. Finally, the results of a comparative case study are presented which uses county-level COVID-19 infection rates and nineteen socio-economic variables from the contiguous United States which demonstrates the merit of Norma’s association analysis approach.

The COVID-19 global pandemic has been a major catastrophic event that impacted the world’s economy. During the pandemic there was a rise in the use of social media such as Twitter by people to express their reactions and responses to the global pandemic. This drove researchers to analyze these micro-blogging texts, using natural language processing (NLP) methods, to understand information inherent in those texts. Most of these NLP tasks employ the use of word embeddings in training neural network models. These word embeddings are mainly trained on general text corpus which produce sub-optimal performance when used in domain-specific NLP tasks such as in COVID-19 related tweets. In this paper, we present a learned COVID-19 tweets domain-specific word embeddings for use in COVID-19 related tweets NLP tasks. Our evaluation results show that our domain-specific COVID-19 tweets word embeddings perform better than pretrained general word embeddings in a downstream domain-specific NLP task. Our COVID-19 tweets word embeddings are available for use by researchers who wish to perform downstream NLP tasks with pretrained domain-specific COVID-19 tweets word embeddings.

Finding the appropriate set of features in cardiac MRI images to localize different areas and anomalies of heart is an essential problem. Convolutional neural networks are known to be well suited for the task to extract features from gray scale images where the intensity is enhanced. This paper proposes a convolutional neural network architecture which can be used to localize the myocardial infarction on the heart using low resolution late gadolinium enhanced (LGE) images. LGE images are T1 weighted MRI images that use contrast agents to increase the intensity of regions where blood accumulates, as a result, areas of heart like ventricle and infarction are brighter as compared to other regions of the heart. We create a U-Net inspired model and train it with the LGE cardiac images to locate and segment the infarction. Our data set has small number of training images with low contrast, which makes it difficult to generalize the features of infarction. To tackle this issue, we propose geometric transformations and pixel intensity manipulations, that should be used for augmenting LGE images to create a diverse training data set. We also propose a chained U-Net approach to reduce the search space for segmenting infarction in LGE cardiac images. Our analysis show a reduction of error from 82% to 68% in segmentation by using the proposed augmentation and the chaining technique. However, it falls short of human level accuracy. Later part of this paper describes the limitation of our current work and lists the future work to overcome those limitations.

This work presents a novel deep learning architecture called BNU-Net for the purpose of cardiac segmentation based on short-axis MRI images. Its name is derived from the Batch Normalized (BN) U-Net architecture for medical image segmentation. New generations of deep neural networks (NN) are called convolutional NN (CNN). CNNs like U-Net have been widely used for image classification tasks. CNNs are supervised training models which are trained to learn hierarchies of features automatically and robustly perform classification. Our architecture consists of an encoding path for feature extraction and a decoding path that enables precise localization. We compare this approach with a parallel approach named U-Net. Both BNU-Net and U-Net are cardiac segmentation approaches: while BNU-Net employs batch normalization to the results of each convolutional layer and applies an exponential linear unit (ELU) approach that operates as activation function, U-Net does not apply batch normalization and is based on Rectified Linear Units (ReLU). The presented work (i) facilitates various image preprocessing techniques, which includes affine transformations and elastic deformations, and (ii) segments the preprocessed images using the new deep learning architecture. We evaluate our approach on a dataset containing 805 MRI images from 45 patients. The experimental results reveal that our approach accomplishes comparable or better performance than other state-of-the-art approaches in terms of the Dice coefficient and the average perpendicular distance.

Nowadays, growing effort has been put to develop spatio-temporal clustering approaches that are capable of discovering interesting patterns in large spatio-temporal data streams. In this paper, we propose a 3-phase serial, density-contour based clustering algorithm called ST-COPOT, which can identify spatio-temporal cluster at multiple levels of density granularity. ST-COPOT takes the point cloud data as input and divides it into batches, next, it employs a non-parametric kernel density estimation approach and contouring algorithms to obtain spatial clusters; at last, spatio-temporal clusters are formed by identifying continuing relationships between spatial clusters in consecutive batches. Moreover, a novel data structure called contour polygon tree is introduced as a compact representation of the spatial clusters obtained for each batch for different density thresholds, and a family of novel distance functions that operate on contour polygon trees are proposed to identify continuing clusters. The experimental results on NYC taxi trips data show that ST-COPOT can effectively discover interesting spatio-temporal patterns in taxi pickup location streams.

This paper focuses on a new type of taxonomy called supervised taxonomy (ST). Supervised taxonomies are generated considering background information concerning class labels in addition to distance metrics, and are capable of capturing class-uniform regions in a dataset. A hierarchical, agglomerative clustering algorithm, called STAXAC that generates STs is proposed and its properties are analyzed. Experimental results are presented that show that STAXAC produces purer taxonomies than the neighbor-joining (NJ) algorithm - a very popular taxonomy generation algorithm. We introduced novel measures and algorithms that assess classification complexity, class modality, and show that STs can be used as the main input of an effective data-editing tool to enhance the accuracy of k-nearest neighbor classifiers. We demonstrated in our experimental evaluation that assessing the classification complexity of a ST provides us with a good estimate of the difficulty of the classification problem at hand. Moreover, a class modality discovery tool (CMD) has been provided that - based on a domain expert's notion of what constitutes a “note-worthy” subclass-determines if specific classes in the dataset are zero-modal, unimodal, and multi-modal.

We define interestingness hotspots as contiguous regions in space which are interesting based on a domain expert's notion of interestingness captured by an interestingness function. This paper centers on finding interestingness hotspots on very large gridded datasets which are quite common in scientific computing. Mining large gridded datasets with a lot of variables and measurements requires a scalable framework that can process large amounts of data in an efficient way. In our recent work, we proposed a computational framework which discovers interestingness hotspots in gridded datasets using a 3-step approach which consists of seeding, hotspot growing and post-processing steps. In this paper, we significantly improve the efficiency of the framework by utilizing parallel processing and employing more efficient data structures and algorithms. We propose a novel heap-based hotspot growing algorithm which brings down the cost of hotspot growing phase significantly. In addition, we propose a graph-based preprocessing algorithm which decreases the number of hotspots grown by merging some hotspot seeds. Other improvements to the framework involve incremental calculation of interestingness functions, and growing hotspots in parallel. The improved framework is evaluated in a case study for a very large 4-dimensional gridded air pollution dataset in which we find interestingness hotspots with respect to pollutants.

This paper addresses two main challenges for clustering which require extensive human effort: selecting appropriate parameters for an arbitrary clustering algorithm and identifying alternative clusters. We propose an architecture and a concrete system MR-CLEVER for multi-run clustering that integrates active learning with clustering algorithms. The key hypothesis of this work is that better clustering results can be obtained by combining clusters that originate from multiple runs of clustering algorithms. By defining states that represent parameter settings of a clustering algorithm, the proposed architecture actively learns a state utility function. The utility of a parameter setting is assessed based on clustering run-time, quality and novelty of the obtained clusters. Furthermore, the utility function plays an important role in guiding the clustering algorithm to seek novel solutions. Cluster novelty measures are introduced for this purpose. Finally, we also contribute a cluster summarization algorithm that assembles a final clustering as a combination of high-quality clusters originating from multiple runs. Merits of our proposed system are that it is generic and therefore can be used in conjunction with different clustering algorithms, and it reduces human effort for selecting the parameters, for comparing clustering results and for assembling clustering results. We evaluate the proposed system in conjunction with a representative based clustering algorithm namely CLEVER for a challenging data mining task involving an earthquake dataset. The obtained results demonstrate that, in comparison to the best single-run clustering, multi-run clustering discovers solutions of higher quality.

Traditional regression analysis derives global relationships between variables and neglects spatial variations in variables. Hence they lack the ability to systematically discover regional relationships and to build better models that use this regional knowledge to obtain higher prediction accuracies. Since most relationships in spatial datasets are regional, there is a great need for regional regression methods that derive regional regression functions that reflect different spatial characteristics of different regions. This paper proposes a novel regional regression framework that first discovers interesting regions showing strong regional relationships between the dependent and the independent variables, and then builds a prediction model with a regional regression function associated with each region. Interesting regions are identified by running a representative-based clustering algorithm that maximizes an externally plugged in fitness function. In this work, we propose two fitness functions: an R-squared based fitness function and an AIC-based fitness function to handle overfitting better. We evaluate our framework in two case studies; (1) identifying causes of arsenic contamination in Texas water wells and (2) Boston Housing dataset determining spatially varying effects of house properties on house prices. We demonstrated that our framework effectively identifies interesting regions and builds better prediction systems that rely on regional models.

This paper proposes a novel framework for mining regional co-location patterns with respect to sets of continuous variables in spatial datasets. The goal is to identify regions in which multiple continuous variables with values from the wings of their statistical distribution are co-located. A co-location mining framework is introduced that operates in the continuous domain and which views regional co-location mining as a clustering problem in which an externally given fitness function has to be maximized. Interestingness of co-location patterns is assessed using products of z-scores of the relevant continuous variables. The proposed framework is evaluated by a domain expert in a case study that analyzes Arsenic contamination in Texas water wells centering on regional co-location patterns. Our approach is able to identify known and unknown regional co-location patterns, and different sets of algorithm parameters lead to the characterization of Arsenic distribution at different scales. Moreover, inconsistent colocation sets are found for regions in South Texas and West Texas that can be clearly attributed to geological differences in the two regions, emphasizing the need for regional co-location mining techniques. Moreover, a novel, prototype-based region discovery algorithm named CLEVER is introduced that uses randomized hill climbing, and searches a variable number of clusters and larger neighborhood sizes.

Efficient means of determining factors controlling spatial distribution of an environmental class variable are of significant interest in Earth science. In this paper, we present a method for automated discovery of controlling factors by mining for emerging patterns in a database constructed from the fusion of several explanatory datasets. We introduce a new definition of pattern support to account for spatial character of the data and systematically evaluate the effectiveness of our technique using a real-world application pertaining to density of vegetation cover. Experimental results show that our method can successfully identify controlling factors for the presence of high vegetation cover.

The basic idea of traditional density estimation is to model the overall point density analytically as the sum of influence functions of data points. However, traditional density estimation techniques only consider the location of a point. Supervised density estimation techniques, on the other hand, additionally consider a variable of interest that is associated with a point. Density in supervised density estimation is measured as the product of an influence function with the variable of interest. Based on this novel idea, a supervised density-based clustering named SCDE is introduced and discussed in detail. The SCDE algorithm forms clusters by associating data points with supervised density attractors which represent maxima and minima of a supervised density function.

The immense explosion of geographically referenced data calls for efficient discovery of spatial knowledge. One of the special challenges for spatial data mining is that information is usually not uniformly distributed in spatial datasets. Consequently, the discovery of regional knowledge is of fundamental importance for spatial data mining. This paper centers on discovering regional association rules in spatial datasets. In particular, we introduce a novel framework to mine regional association rules relying on a given class structure. A reward-based regional discovery methodology is introduced, and a divisive, grid-based supervised clustering algorithm is presented that identifies interesting subregions in spatial datasets. Then, an integrated approach is discussed to systematically mine regional rules. The proposed framework is evaluated in a real-world case study that identifies spatial risk patterns from arsenic in the Texas water supply.

Adaptive clustering uses external feedback to improve cluster quality; past experience serves to speed up execution time. An adaptive clustering environment is proposed that uses Q-learning to learn the reward values of successive data clusterings. Adaptive clustering supports the reuse of clusterings by memorizing what worked well in the past. It has the capability of exploring multiple paths in parallel when searching for good clusters. In a case study, we apply adaptive clustering to instance-based learning relying on a distance function modification approach. A distance function adaptation scheme that uses external feedback is proposed and compared with other distance function learning approaches. Experimental results indicate that the use of adaptive clustering leads to significant improvements of instance-based learning techniques, such as k-nearest neighbor classifiers. Moreover, as a by-product a new instance-based learning technique is introduced that classifies examples by solely using cluster representatives; this technique shows high promise in our experimental evaluation.

The objective of a general-purpose content-based image retrieval system is to find images in a database that match an external measure of relevance. Since users follow different and inconsistent relevance measures, processing queries in a task-specific manner has shown to be an effective approach. Viewing specialized image retrieval algorithms as agents, we propose a general-purpose image retrieval system that uses a new multi-agent meta-learning framework. The framework adapts a distance function defined over both image distance weights and image queries to identify clusters of algorithms that produce similar solutions to similar problems. Experiments compare our approach with a traditional information retrieval algorithm; results show that our framework provides better average relevance scores

The goal of dataset editing in instance-based learning is to remove objects from a training set in order to increase the accuracy of a classifier. For example, Wilson editing removes training examples that are misclassified by a nearest neighbor classifier so as to smooth the shape of the resulting decision boundaries. This paper revolves around the use of representative-based clustering algorithms for nearest neighbor dataset editing. We term this approach supervised clustering editing. The main idea is to replace a dataset by a set of cluster prototypes. A clustering approach called supervised clustering is introduced for this purpose. Our empirical evaluation using eight UCI datasets shows that both Wilson and supervised clustering editing improve accuracy on more than 50% of the datasets tested. However, supervised clustering editing achieves four times higher compression rates than Wilson editing.

This work centers on a novel data mining technique we term supervised clustering. Unlike traditional clustering, supervised clustering assumes that the examples are classified and has the goal of identifying class-uniform clusters that have high probability densities. Four representative-based algorithms for supervised clustering are introduced: a greedy algorithm with random restart, named SRIDHCR, that seeks for solutions by inserting and removing single objects from the current solution, SPAM (a variation of the clustering algorithm PAM), an evolutionary computing algorithm named SCEC, and a fast medoid-based top-down splitting algorithm, named TDS. The four algorithms were evaluated using a benchmark consisting of four UCI machine learning data sets. In general, it seems that "greedy" algorithms, such as SPAM, SRIDHCR, and TDS, do not perform particularly well for supervised clustering and seem to terminate prematurely too often. We also briefly describe the applications of supervised clustering.

We propose a preprocessing step to classification that applies a clustering algorithm to the training set to discover local patterns in the attribute or input space. We demonstrate how this knowledge can be exploited to enhance the predictive accuracy of simple classifiers. Our focus is mainly on classifiers characterized by high bias but low variance (e.g., linear classifiers); these classifiers experience difficulty in delineating class boundaries over the input space when a class distributes in complex ways. Decomposing classes into clusters makes the new class distribution easier to approximate and provides a viable way to reduce bias while limiting the growth in variance. Experimental results on real-world domains show an advantage in predictive accuracy when clustering is used as a preprocessing step to classification.

Factoring is believed to be a difficult task. Although factoring is of interest in its own right, the security of RSA cryptography, among other cryptographic systems, is dependent on the difficulty of factoring the product of large primes. We show how to cast the problem of factoring integers as a state-based search to which the techniques of AI may be applied. Using small primes as operators, goal states can be characterized by integers close to a multiple kN of N, where N is the number to be factored. For a given value of k we formulate a heuristic formula for the resulting search which is universally optimistic. From a basic platform of depth first search we then make modifications, both with standard AI techniques such as pruning and with techniques which are specific for the particular problem. In the process we note the properties of a novel sub-class of integers, pseudo-smooth integers, which are useful in construction of an expanded search. We review empirical evidence of the improvements our various modifications make. Finally, we extend the factoring tool in various ways to deal effectively with the search task when the factor base is relatively small.

One of the most important parts of designing an expert system is elicitation of the expert's knowledge. This knowledge usually consists of facts and rules. Eliciting these rules and facts is relatively easy: the more complicated task is assigning weights (numerical or interval-valued degrees of belief) to different statements from the knowledge base. Experts often cannot quantify their degrees of belief, but they can order them (by suggesting which statements are more reliable). It is, therefore, reasonable to try to reconstruct the degrees of belief from such an ordering.In this paper, we analyze when such a reconstruction is possible, whether it lead to unique values of degrees of belief, and how computationally complicated the corresponding reconstruction problem can be. © 1998 John Wiley & Sons, Inc.

Methodologies based on an elaboration of the knowledge acquisition and design structuring (KADS) philosophy were developed and a suite of tools based on this methodology was implemented. The tools implement object oriented support at the domain-layer, Bayesian reasoning combined with a Bayesian compatible version of "fuzzy sets" to support the inference-layer, and mechanisms for knowledge level task control to support both the task-level and strategic-level. These tools were built for the task of computerizing medical practice guidelines. The tools were successfully applied to two practice guidelines, one selected by symptom (febrile neutropenia) and the other by drug class (CSF).

Whenever a user or an application updates the contents of a database or knowledge base, it has to be guaranteed that the DB or KB is left in a consistent and meaningful state. The admissibility of a state is defined by means of consistency constraints (or semantic integrity constraints). For static constraints, which are defined on a single database state, J.M. Nicolas (1982) proposed an enforcement method in which constraints are defined in first order logic, and the enforcement procedure is efficient. Dynamic constraints concern several database states, associated to possibly distinct time points. They allow us to impose restrictions on the database transitions from a state to another: for example, in personnel management, to ensure that salaries are never decreased; or to monitor instalments for a sale made by a warehouse. The semantics of dynamic constraints is complex because of the temporal dimension. We propose a method for constraints definition and enforcement which is able to deal with dynamic constraints, introducing a distinction between imperative and non imperative constraints. A temporal first order logic is used for constraints definition and the enforcement procedure is extended through the use of active rules. The method extends Nicolas' proposal in that it is able to cope with the static constraints handled by Nicolas' method.

The paper describes an inductive learning environment called DELVAUX for classification tasks that learns PROSPECTOR-style, Bayesian rules from sets of examples. A genetic algorithm approach is used for learning Bayesian rule-sets, in which a population consists of sets of rule-sets that generate offspring through the exchange of rules, permitting fitter rule-sets to produce offspring with a higher probability. To deal with the premature convergence problem, fuzzy similarity measures for Bayesian rule-sets are introduced and the genetic algorithm approach is modified, so that similar rule-sets produce offspring with a lower probability, relying on a sharing function approach. Empirical results are presented that evaluate the benefits of the sharing function approach in our learning environment.>

The paper surveys the features of an inductive learning environment named DELVAUX that learns prospector.

A rule-based approach for the automatic enforcement of consistency constraints is presented. In contrast to existing approaches that compile consistency checks into application programs, the approach centralizes consistency enforcement in a separate module called a knowledge-base management system. Exception handlers for constraint violations are represented as rule entities in the knowledge base. For this purpose, a new form of production rule called the activation pattern controlled rule is introduced: in contrast to classical forward chaining schemes, activation pattern controlled rules are triggered by the intent to apply a specific operation but not necessarily by the result of applying this operation. Techniques for implementing this approach are discussed, and experiments in speeding up the system performance are described. Furthermore, an argument is made for more tolerant consistency enforcement strategies, and how they can be integrated into the rule-based approach to consistency enforcement is discussed.>

The human capability of making complex decisions is one of the most fascinating facets of human intelligence, especially if vague, judgemental, default, or uncertain knowledge is involved. This paper focuses on the automation of this capability. It advocates treating rule-based decision making as a problem of evidence combination: Rules are considered to provide evidence for or against a certain decision; after the rules have been processed, the evidence associated with the individual decisions is combined and the best decision is selected. Advantages and problems of this approach are discussed and experiences the authors had in applying this approach in a Computer Bridge project are surveyed. Finally, a language construct, called a decision block, that facilitates the automation of rule-based decision making involving imperfect knowledge is introduced. The benefits of decision blocks for writing complex decision support systems are evaluated, and their integration with a rule-based forward chaining language is discussed.>

As part of developing the CLIPS-based debugging tool MIRO, and because there has been relatively little exploration of the debugging process for such programs, the problems of debugging data-driven rule-based programs in general are defined and analyzed. How explanation can be used as a tool to ease the debugging process is discussed, and how it can suggest causes of faults in answer to questions asking why a particular fault occurred is examined. An algorithm that provides part of the explanation for a rule's failure to fire is discussed.>

To debug a forward-chaining rule-based program, certain, historical, information is needed. System builders should be able to request such information directly, instead of having to rerun the program one step at a time or search a trace of run details. As a first step in designing an explanation system for answering such questions, a proposal is discussed for storing a forward-chaining program run's 'historical' details in its Rete inference network, used to match rule conditions to working memory, without seriously affecting the network's run-time performance. This proposed modified Rete network is called a historical Rete network. Various algorithms for maintaining this network are discussed, along with how it can be used to analyze what happened during a program run.>

Article Free Access Share on Acquisition of terminological knowledge using database design techniques Authors: Christoph F. Eick Fakultat fur Informatik, Universitat Karlsruhe, Postfach 6380, D-7500 Karlsruhe Fakultat fur Informatik, Universitat Karlsruhe, Postfach 6380, D-7500 KarlsruheView Profile , Peter C. Lockemann Fakultat fur Informatik, Universitat Karlsruhe, Postfach 6380, D-7500 Karlsruhe Fakultat fur Informatik, Universitat Karlsruhe, Postfach 6380, D-7500 KarlsruheView Profile Authors Info & Claims SIGMOD '85: Proceedings of the 1985 ACM SIGMOD international conference on Management of dataMay 1985Pages 84–94https://doi.org/10.1145/318898.318905Published:01 May 1985Publication History 20citation379DownloadsMetricsTotal Citations20Total Downloads379Last 12 Months14Last 6 weeks2 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteeReaderPDF

A comprehensive and constructive design methodology for Logical data base design is proposed. A set of compatible computerized design tools is described, carrying out the following tasks: formalization of natural language information requirements; refinement, evaluation and transformation of conceptual data definitions; mapping conceptual to DBTG data definitions.