Continuous Attributes Research Articles

Improve the Classifier Accuracy for Continuous Attributes in Biomedical Datasets Using a New Discretization Method

In real-time data mining applications discrete values play vital role in knowledge representation as they are easy to handle and very close to knowledge level representation than continuous attributes. Discretization is a major step in data mining process where continuous attributes are transformed into discrete values. However, most of the classifications algorithms are require discrete values as the input. Even though some data mining algorithms directly contract with continuous attributes, the learning process yields low quality results. In this paper, we introduce a new discretization method based on standard deviation technique called ‘z-score’ for continuous attributes on biomedical datasets. We compare performance of the proposed algorithm with the state-of- the-art discretization techniques. The experiment results show the efficiency in terms of accuracy and also minimize the classifier confusion for decision making process.

Drawing inferences from clinical studies with missing values using genetic algorithm.

Missing data problem degrades the statistical power of any analysis made in clinical studies. To infer valid results from such studies, suitable method is required to replace the missing values. There is no method which can be universally applicable for handling missing values and the main objective of this paper is to introduce a common method applicable in all cases of missing data. In this paper, Bayesian Genetic Algorithm (BGA) is proposed to effectively impute both missing continuous and discrete values using heuristic search algorithm called genetic algorithm and Bayesian rule. BGA is applied to impute missing values in a real cancer dataset under Missing At Random (MAR) and Missing Completely At Random (MCAR) conditions. For both discrete and continuous attributes, the results show better classification accuracy and RMSE% than many existing methods.

Extracting a cancer model by enhanced ant colony optimisation algorithm.

Although Ant-Miner has been used with relative ease for datasets with categorical data and small-sized feature vectors, microarray datasets, which contain a few samples with large amount of genes, are a totally different story. The Ant-Miner is an ant colony optimisation algorithm that extracts predictive rules from datasets and intrinsically works on discrete values. This study has developed a new algorithm, "Enhanced Ant-Miner" (EAM), based on previous works. EAM deals with continuous attributes as well as categorical ones and presents its captured models in the form of predictive rules. EAM has been tested versus SVM, CN2, K-means and hierarchical clustering and the results show that EAM is the best in the context of predictive accuracy. Additionally, its agent-based nature gives it a much more charming ability to speed up the whole process when compared to other trivial miners.

A Novel Discretization Method for Continuous Attributes: A Machine Learning Approach

Discretization plays a vital role where continuous attributes are transformed into discrete values for preprocessing tasks in data mining algorithms and machine learning classifiers. Most of the real data often come in mixed notations, i.e., continuous and discrete, while many machine learning algorithms require it in the form of discrete data only. In the past few decades, numerous techniques have been proposed as discretization techniques for continuous data, which are unable to provide better accuracy, understandability of models and also classifier confusion in the form of continuous data. In this paper, we propose a new discretization technique called ‘ZDisc’ based on the standard deviation normalization statistical technique for continuous attributes to generate less number of cut points. The proposed method has been compared with other state-of-the-art discretization techniques on benchmark continuous datasets. The experiment shows that the proposed discretization method performs competitive discretization schemes in terms of classifier accuracy and to minimize the classifi er confusion.

Pufferfish

In this article, we introduce a new and general privacy framework called Pufferfish. The Pufferfish framework can be used to create new privacy definitions that are customized to the needs of a given application. The goal of Pufferfish is to allow experts in an application domain, who frequently do not have expertise in privacy, to develop rigorous privacy definitions for their data sharing needs. In addition to this, the Pufferfish framework can also be used to study existing privacy definitions. We illustrate the benefits with several applications of this privacy framework: we use it to analyze differential privacy and formalize a connection to attackers who believe that the data records are independent; we use it to create a privacy definition called hedging privacy, which can be used to rule out attackers whose prior beliefs are inconsistent with the data; we use the framework to define and study the notion of composition in a broader context than before; we show how to apply the framework to protect unbounded continuous attributes and aggregate information; and we show how to use the framework to rigorously account for prior data releases.

Fuzzy Rough Decision Trees

How to evaluate features and select nodes is one of the key issues in constructing decision trees. In this work fuzzy rough set theory is employed to design an index for evaluating the quality of fuzzy features or numerical attributes. A fuzzy rough decision tree algorithm, which can be used to address classification problems described with symbolic, real-valued or fuzzy features, is developed. As node selection, split generation and stopping criterion are three main factors in constructing a decision tree, we design different techniques to determine splits with different kinds of features. The proposed algorithm can directly generate a classification tree without discretization or fuzzification of continuous attributes. Some numerical experiments are conducted and the comparative results show that the proposed algorithm is effective compared with some popular algorithms.

퍼지신경망을 사용한 네이브 베이지안 분류기의 분산 그래프 학습

Abstract Naive Bayesian classifiers are a powerful and well-known type of classifiers that can be easily induced from a dataset of sample cases. However, the strong conditional independence assumptions can sometimes lead to weak classification performance. Normally, naive Bayesian classifiers use Gaussian distributions to handle continuous attributes and to represent the likelihood of the features conditioned on the classes. The probability density of attributes, however, is not always well fitted by a Gaussian distribution. Another eminent type of classifier is the neuro-fuzzy classifier, which can learn fuzzy rules and fuzzy sets using supervised learning. Since there are specific structural similarities between a neuro-fuzzy classifier and a naive Bayesian classifier, the purpose of this study is to apply learning distribution graphs constructed by a neuro-fuzzy network to naive Bayesian classifiers. We compare the Gaussian distribution graphs with the fuzzy distribution graphs for the naive Bayesian classifier. We applied these two types of distribution graphs to classify leukemia and colon DNA microarray data sets. The results demonstrate that a naive Bayesian classifier with fuzzy distribution graphs is more reliable than that with Gaussian distribution graphs.

Fast Fuzzy Search for Mixed Data Using Locality Sensitive Hashing

The drastic increase in data volume strongly demands efficient search techniques for similar data to queries. It is sometimes useful to specify data of interest with fuzzy constraints. When data objects contain both numerical and categorical attributes, it is usually not easy to define commonly-accepted distance measures between data objects. With no efficient indexing structure, it costs much to search for specific data objects because a linear search needs to be conducted over the whole data set. This paper proposes a method to use locality sensitive hashing technique and fuzzy constrained queries to search for interesting ones from big data. The method builds up a locality sensitive hashing-based indexing structure only with constituting continuous attributes, collects a small number of candidate data objects to which query is examined, and then evaluates their satisfaction degree to fuzzy constrained query so that data objects satisfying the query are determined.

Interval Similarity-Based Quantization Method for Continuous Data

Data quantization methods for continuous attributes play an extremely important role in artificial intelligence, data mining and machine learning because discrete values of attributes are required in most classification methods. In this paper, we present an interval similarity-based quantization method for continuous data. It defines an interval similarity criterion which is regarded as a new merging standard in the process of quantization. In addition, a heuristic quantization algorithm is proposed to achieve a satisfying quantization result with the aim to improve the performance of inductive learning algorithms. The new algorithm realizes fair standard and quantifying the real value attributes exactly and reasonably. Empirical experiments on UCI real data sets show that our proposed algorithm generates a better quantization scheme that improves the classification accuracy of inductive learning than existing quantization algorithms.

Rough Sets Algorithm and Its Application in Fault Diagnosis

Gearbox is one of the most complicate rotary mechanical apparatus, the fault signal shows non-linear and non-stationary, and how to recognize the faults effectively is a key issue. A novel method based on wavelet packet transform and rough sets theory was presented for fault diagnosis of gearbox. First, the vibration signals were decomposed into eight bands from low frequency to high frequency by wavelet packet transform, energy characteristics were extracted as the condition attributes. Second, an improved NaiveScaler algorithm was put forward to discrete continuous attributes in the case of assuring classification ability. A new reduction algorithm based on condition equivalence classifications was proposed to delete the redundant features, which could improve the reduction efficiency. Lastly the decision rules were drawn and utilized to test the samples. The results show that the method could obtain more sensitive fault characteristic parameters and have better classification ability accordingly. DOI: http://dx.doi.org/10.11591/telkomnika.v11i9.3308

Application of Variable Precision Rough Set and Integrated Neural Network to Bearing Fault Diagnosis

The integration of variable precision rough set and neural network is introduced into the bearing fault diagnosis. VPRS-INN fault diagnosis method is proposed: First, utilize the information entropy method for discretization of continuous attributes, and then use attribute dependence degree of the variable precision rough set theory for heuristic reduction. based on the reduction, obtain the optimal decision support system. Finally according to the optimal design system, we design a integrated neural network for fault diagnosis. instances have proved the feasibility and high fault diagnosis rate of the method.

A Supervised Statistical Data Quantization Method in Machine Learning

Data quantization methods for continuous attributes play an extremely important role in artificial intelligence, data mining and machine learning because discrete values of attributes are required in most classification methods. In this paper, we present a supervised statistical data quantization method. It defines a quantization criterion based on the chi-square statistic to discover accurate merging intervals. In addition, a heuristic quantization algorithm is proposed to achieve a satisfying quantization result with the aim to improve the performance of inductive learning algorithms. Empirical experiments on UCI real data sets show that our proposed algorithm generates a better quantization scheme that improves the classification accuracy of C4.5 decision tree than existing algorithms

Comparing discretization and selection methods for the logical-combinatorial classification of continuous parameters

Preprocessing methods for handling problems with features containing continuous attributes are discussed for learning a classification algorithm based on the JSM method. Discretization methods for continuous parameters that do not make use of class information on feature distribution are compared to entropy-based methods employing class labels in interval partitioning. An entropy-information-based method for selecting attributes is also discussed.

Continuous Attributes Discretization Algorithm based on FPGA

The paper addresses the problem of Discretization of continuous attributes in rough set. Discretization of continuous attributes is an important part of rough set theory because most of data that we usually gain are continuous data. In order to improve processing speed of discretization, we propose a FPGA-based discretization algorithm of continuous attributes making use of the speed advantage of FPGA. Combined attributes dependency degree of rough ret, the discretization system was divided into eight modules according to block design. This method can save much time of pretreatment in rough set and improve operation efficiency. Extensive experiments on a certain fighter fault diagnosis validate the effectiveness of the algorithm. DOI: http://dx.doi.org/10.11591/telkomnika.v11i7.2811

A Novel Method for Mining Association Rules from Continuous Attributes Based on Cultural Immune Algorithm

It is an important topic to mine association rules from data with continuous attributes. A novel model based on cultural algorithm and immune algorithm is proposed in this paper. This algorithm uses the cultural algorithm framework in which the immune algorithm embedded. The immune algorithm is used to discretize the continuous attributes and search association rules in the database rapidly. The cultural algorithm is used to obtain the commonly accepted beliefs in order to guide and speed up the search. The new algorithm integrates the discretization, attributes reduction and association rules mining together. In addition, a new diversity operator is put forward in order to discover global solutions. The experiment shows that the new algorithm is superior to immune algorithm in convergence speed and the rule’s accuracy.

A Discretization Algorithm Based on Cultural Immune Algorithm

Now more and more researchers pay attention to discretization of the continuous attributes. A hybrid evolutionary system based on immune algorithm and cultural algorithm is proposed for discretization which combines the strengths of evolutionary computing, social computing in this paper. It uses the cultural algorithm framework in which the immune algorithm embedded. Immune algorithm discretizes the continuous data and the cultural algorithm constitutes the commonly accepted beliefs to guide and speed up the search. In addition, a new diversity operator is put forward in order to maintain the diversity of population. Finally, experiments are conducted for extensive performance testing of the new method on several publicly available data sets. The results show that the algorithm has higher convergence speed and less classification errors than many previously known discretization methods.

Review on Application of Rough Set Theory

Rough set theory has found an increasingly wide utilization since it was promoted in 1980s.And study on the application of rough set theory in every field has a great development in recent years. Application of rough set theory in attribute reduction, continuous attributes discretization, and uncertainty measuring, as well as application of information entropy in rough set theory are reviewed in this paper. What we will do next is to probe further into the application of information entropy in rough set theory.

A Survey of Discretization Techniques: Taxonomy and Empirical Analysis in Supervised Learning

Discretization is an essential preprocessing technique used in many knowledge discovery and data mining tasks. Its main goal is to transform a set of continuous attributes into discrete ones, by associating categorical values to intervals and thus transforming quantitative data into qualitative data. In this manner, symbolic data mining algorithms can be applied over continuous data and the representation of information is simplified, making it more concise and specific. The literature provides numerous proposals of discretization and some attempts to categorize them into a taxonomy can be found. However, in previous papers, there is a lack of consensus in the definition of the properties and no formal categorization has been established yet, which may be confusing for practitioners. Furthermore, only a small set of discretizers have been widely considered, while many other methods have gone unnoticed. With the intention of alleviating these problems, this paper provides a survey of discretization methods proposed in the literature from a theoretical and empirical perspective. From the theoretical perspective, we develop a taxonomy based on the main properties pointed out in previous research, unifying the notation and including all the known methods up to date. Empirically, we conduct an experimental study in supervised classification involving the most representative and newest discretizers, different types of classifiers, and a large number of data sets. The results of their performances measured in terms of accuracy, number of intervals, and inconsistency have been verified by means of nonparametric statistical tests. Additionally, a set of discretizers are highlighted as the best performing ones.

A hybrid feature selection scheme for mixed attributes data

Feature selection aims at reducing the number of features in many applications. Existing feature selection approaches mainly deals with classification problems with continuous or discrete attributes. However, data usually come with mixed attributes in real-world applications. In this paper, a hybrid feature selection (HFS) scheme is proposed to deal with mixed attributes data. Firstly, a new correlation measure between mixed attributes is defined by giving a model for calculating mutual information between continuous and discrete attributes; secondly, the features are evaluated by a filter model with the new correlation measure; finally, feature selection is done by optimizing the parameter in the filter model with estimation accuracy criterion. Experimental results show that HFS acquires better stability and estimation accuracy.

Non-Naive Bayesian Classifiers for Classification Problems With Continuous Attributes

An important way to improve the performance of naive Bayesian classifiers (NBCs) is to remove or relax the fundamental assumption of independence among the attributes, which usually results in an estimation of joint probability density function (p.d.f.) instead of the estimation of marginal p.d.f. in the NBC design. This paper proposes a non-naive Bayesian classifier (NNBC) in which the independence assumption is removed and the marginal p.d.f. estimation is replaced by the joint p.d.f. estimation. A new technique of estimating the class-conditional p.d.f. based on the optimal bandwidth selection, which is the crucial part of the joint p.d.f. estimation, is applied in our NNBC. Three well-known indexes for measuring the performance of Bayesian classifiers, which are classification accuracy, area under receiver operating characteristic curve, and probability mean square error, are adopted to conduct a comparison among the four Bayesian models, i.e., normal naive Bayesian, flexible naive Bayesian (FNB), the homologous model of FNB (FNBROT), and our proposed NNBC. The comparative results show that NNBC is statistically superior to the other three models regarding the three indexes. And, in the comparison with support vector machine and four boosting-based classification methods, NNBC achieves a relatively favorable classification accuracy while significantly reducing the training time.