Classifying Multiple Imbalanced Attributes in Relational Data

XML (Extensible Markup Language) has emerged as the standard data-exchange format for Internet-based business applications. These applications introduce a new set of data management requirements involving XML. However, for the foreseeable future, a significant amount of business data will continue to be stored in relational database systems. Thus, a bridge is needed to satisfy the requirements of these new XML-based applications while still using relational database technology. This paper describes the design and implementation of the XTABLES middleware system, which we believe achieves this goal. In particular, XTABLES provides a general framework to create XML views of relational data, query XML views, and store and query XML documents using a relational database system. Some of the novel features of the XTABLES architecture are that it 1) provides users with a single XML query language for creating and querying XML views of relational data, 2) executes queries efficiently by pushing most computation down to the relational database engine, c) allows users to query seamlessly over relational data and meta-data, and 4) allows users to write queries that span XML documents and XML views of relational data.

The Next Generation Linkage Management System (NGLMS) was designed around keeping all data in a graph database. However, this constraint, while easily achievable for greenfield projects and/or new data linkage units, may not be easily met where legacy data exists.
 Objectives and ApproachThe NGLMS was extended to encompassed system where data was held partially or even completely in a relational database. By grouping the data managed by the NGLMS into system metadata, record link data and record data and allowing system metadata and record data to be stored separately and independently in either a relational or a graph database, the NGLMS allows hybrid installations of mixed graph and relational data and, with some loss of functionality, purely relational installations.
 ResultsThe functionality of the NGLMS was expanded to allow review of existing legacy data stored in a relational database system. Through minor changes to the server used by the NGLMS Clerical Review tool (NGLMS-CR) the review tools was able to present the same interface and allow the same integration as project stored completely within the graph database.
 Hybrid projects where link information was stored in a graph could be accommodated with no loss of functionality.Relational-only projects allowed clerical review of identified clusters in a manner identical to the graph-only NGLMS but involved some curtailment of the advanced clustering, time-slicing, compositional and concurrent functionality of the NGLMS due to the loss of the functionality provided by a graph database. They do provide an upgrade pathway to hybrid projects and then graph-only projects.
 Conclusion / ImplicationsAllowing the NGLMS to be configured as a hybrid system enables a gradual adoption of the NGLMS toolset and software.
 
 purely relational data still allows the use of the NGLMS-CR with customisable workpools and workflows
 a hybrid relational/graph system where record data is still kept in a relational store but cluster and linkage information is kept in the graph store allows the use of legacy data without major disruption and provides a pathway to full adoption

Reasoning about attribute value equivalence in relational data

How can we analyze large-scale real-world data with various attributes? Many real-world data (e.g., network traffic logs, web data, social networks, knowledge bases, and sensor streams) with multiple attributes are represented as multi-dimensional arrays, called tensors. For analyzing a tensor, tensor decompositions are widely used in many data mining applications: detecting malicious attackers in network traffic logs (with source IP, destination IP, port-number, timestamp), finding telemarketers in a phone call history (with sender, receiver, date), and identifying interesting concepts in a knowledge base (with subject, object, relation). However, current tensor decomposition methods do not scale to large and sparse real-world tensors with millions of rows and columns and `fibers.' In this paper, we propose HaTen2, a distributed method for large-scale tensor decompositions that runs on the MapReduce framework. Our careful design and implementation of HaTen2 dramatically reduce the size of intermediate data and the number of jobs leading to achieve high scalability compared with the state-of-the-art method. Thanks to HaTen2, we analyze big real-world sparse tensors that cannot be handled by the current state of the art, and discover hidden concepts.

A Personalized Recommender System from Probabilistic Relational Model and Users’ Preferences

The RDF framework is becoming popular for presenting data. It makes the data easily accessible and queryable. However, the most common way how to store structured data is to use a relational database system. It is essential to create a mapping between these two worlds, to publish the data stored in a relational database in the RDF format. That can be effectively achieved by a virtual SPARQL endpoint over relational data.

Relational database system (RDBMS-Relational Database Management System) is a database system that has a very strong mathematical foundation and has many advantages, so it is generally accepted by almost vendor. Basically, there is nothing 'wrong' with a relational database system. However, with the emergence of a variety of Web applications today that require the higher performance of query, relational data base system was showing the queries performance that do not fit the user’s needs and expectations, because relational database system, for generate specific queries, often have to join data from two or more tables that relate. This process require relational database server to perform a data search (lookup) in the tables must be joined. These data searches require computing resources (processor performance and memory usage) that relatively high, so the performance of a relational database server typically decreasesas more data should be joined. This problems trying to be addressed by several vendor. Google with its BigTable, Amazon with its Dynamo, Yahoowith its PNUTS, and Facebook with its Apache Cassandra for example, using the non-relationaldata storage (Data Store) approach. The purpose of this paper is to introduce Apache Cassandra that is one of the very large non-relational systems, then compare to relational database systems.

This research work proposes a Weighted Least Square Twin Support Vector Machine (WLSTSVM) for imbalanced dataset. Real world data are imbalanced in nature due to which most of the classification techniques do not work well. In Imbalanced data, there is a huge difference between the numbers of data samples of classes. One class data samples are larger as compared to other class data samples. This paper discusses the traditional methods of handling imbalanced data and proposes an improvement over Least Square Twin Support Vector Machine. This research work has performed experiment on five benchmark UCI datasets using 10-fold cross validation method. The results of experiment show that the proposed technique performed well for imbalanced dataset and its accuracy is better as compared to other existing methods. This research work presents the formulation of proposed approach for both linear and non-linear data samples.

Power distribution systems reliability is significantly affected by many outage causing events; good outage cause identification can help expedite the restoration procedure. However, the data imbalance issue encountered in many real-world data affects the performance of fault cause identification. The elegant fuzzy classification algorithm, I-algorithm, proposed by Ishibuchi et al. achieves satisfactory performance on many carefully preprocessed data sets but not on the imbalanced data, I-algorithm, an extension of the I-algorithm, is developed in this paper to alleviate the effect of imbalanced data constitution. Both the I- and E-algorithms are applied to Duke Energy outage data for power distribution systems fault cause identification. Their performance on this real-world imbalanced data set is presented, compared, and analyzed to demonstrate the improvement achieved by the extended algorithm

In day today life, data is generated in massive amount with rapid growth in health care environment. The medical industries have large amount of data sets, for diagnosis purpose and maintain patient's records. The medical researches come with new treatments and medicine every day. But availability of medical datasets is often not balanced in their class labels. The performance of some existing method is poor on imbalanced dataset. So the prediction of disease from imbalanced data becomes difficult to handle. In this proposal Classifier ensemble method (Random Forest algorithm) can be used to overcome existing classifier techniques. Multiple classifier system is more accurate and robust than an existing classifier technique. The ensemble method proves to be very efficient in classification of records from available imbalanced healthcare patient data, as it involves the process of considering opinion from multiple base classifiers, as opposed to the single classifier method. This method gives a very accurate and precise inference, as unrelated data's are removed because of multiple base classifiers. The problems of healthcare dataset especially with some uncertainty can be predicted.

The future of computer aided design is in object oriented programming. If the database community hopes to participate in this future, it must reexamine some basic assumptions about the architecture of database systems. Database system functionality can be added to object systems but if the performance cost is too high, it will never survive. Below are some suggestions for what can be done at a reasonable performance cost. The object oriented paradigm provides a more practical approach to the partitioning of the global database than horizontal and vertical partitioning of relational tables. Each partition should itself be an independent database containing related data such as the geometry of a part or the spacial relationship of parts in an assembly. A meta-database would be used to control access to collections of these partitions. A collection of partitions comprise the database for a user's design session. The overhead of traditional database transaction management is not acceptable for high performance CAD systems. With the partitioning scheme described above, transaction management can be performed at a partition/session granularity. Once the user has composed the collection of partitions, he has a single user database. There is no need for concurrency control or transaction logging except at the meta-database level. This type of transaction management can in fact be more functional than traditional transaction management, allowing for versioning, long transactions, integrity checking and archival. Object oriented databases need a message model, not a data model. Any object which responds to the same messages as an object of “Duck” class (walk and quack) is, for all intents and purposes, a duck. An attempt to design a data model based on instance variables of an object or based on collections of objects of like class violates the data abstraction facilities of object oriented languages and diminishes their power. An attempt to implement a relational database system with an object oriented language yields a relational database system where you get abstract data types for free. It does not yield an object oriented database system. For object oriented queries, the message is the media. A query can be transformed into an execution plan consisting of messages sent to database objects. Optimization decisions can be made by sending messages to referenced objects. Collection classes can be implemented for new access methods with cost and selectivity methods to provide optimization information. In this way, the query language can grow with the application. Data representation is an important aspect of object oriented systems. Most object systems are typeless in that all instance variables of an object are object references. For performance sake, object systems should provide enough of a type mechanism to allow simple data items (integers, floats, characters … ) to be represented in the form intrinsic to the machine. Methods can then be compiled for access to typed data. In conclusion, object systems provide enormous potential for the development of CAD systems. Performance influences the approach taken to an application. WYSIWYG publishing applications were not attempted until performance was adequate. Functionality is what sells CAD systems. Database system functionality can be added to object systems at a reasonable cost.

Statistical inference of travelers’ route choice preferences with system-level data

Although significant progress has been made in enhancing visibility, retrieving texture details, and mitigating noise in Low-Light (LL) images, the challenge persists in applying current Low-Light Image Enhancement (LLIE) methods to real-world scenarios, primarily due to the diverse illumination conditions encountered. Furthermore, the quest for generating enhancements that are visually realistic and attractive remains an underexplored realm. In response to these challenges, we present a novel LLIE framework with the guidance of Generative Perceptual Priors (GPP-LLIE) derived from vision-language models (VLMs). Specifically, we first propose a pipeline that guides VLMs to assess multiple visual attributes of the LL image and quantify the assessment to output the global and local perceptual priors. Subsequently, to incorporate these generative perceptual priors to benefit LLIE, we introduce a transformer-based backbone in the diffusion process, and develop a new layer normalization (GPP-LN) and an attention mechanism (LPP-Attn) guided by global and local perceptual priors. Extensive experiments demonstrate that our model outperforms current SOTA methods on paired LL datasets and exhibits superior generalization on real-world data.

Deep learning methods are widely used in the field of rolling bearing fault diagnosis and produce good results when faced with datasets with roughly equal numbers of normal and faulty samples. However, real-world data often has a serious imbalance, with the number of fault samples being significantly less than the number of normal samples. This dataset imbalance challenges the performance of traditional deep learning methods. To address this problem, this paper proposes an efficient imbalanced data rolling bearing fault diagnosis method. The method consists of two parts: a deep learning network based on a multi-scale self-attention mechanism and a novel loss function. In terms of the deep learning network, firstly, the one-dimensional vibration signal is converted into a two-dimensional image through the Gramian angular field. This conversion maximises the inherent feature extraction capability of the network. Subsequently, the multi-scale learning capability of the network is enhanced by implementing different expansion rates for the head of the multi-scale self-attention mechanism. This nuanced approach allows the network to capture the underlying information more efficiently. Finally, the inclusion of Ghost bottlenecks and feature pyramid networks (FPNs) helps to optimise network efficiency and improve generalisation performance. A novel loss function is also proposed to make the method more suitable for imbalanced data. During the training process, the classification of samples in each class is analysed using the recall metric of imbalanced classification and the real-time recall is used as a weight to weaken the dominance of the majority class. This weighting ensures the adaptability of the method to imbalanced datasets. The proposed method is evaluated using rolling bearing datasets from Case Western Reserve University, USA, and Southeast University, China. Comparison results with other state-of-the-art deep learning methods show that the proposed method has a robust performance when dealing with imbalanced data.

Prediction of body mass index status from voice signals based on machine learning for automated medical applications