Xiuyi Jia

Three-Way Decisions Solution to Filter Spam Email: An Empirical Study

A three-way decisions solution based on Bayesian decision theory for filtering spam emails is examined in this paper. Compared to existed filtering systems, the spam filtering is no longer viewed as a binary classification problem. Each incoming email is accepted as a legitimate or rejected as a spam or undecided as a further-exam email by considering the misclassification cost. The three-way decisions solution for spam filtering can reduce the error rate of classifying a legitimate email to spam, and provide a more meaningful decision procedure for users. The solution is not restricted to a specific classifier. Experimental results on several corpus show that the three-way decisions solution can get a better total cost ratio value and a lower weighted error.

Sentiment Analysis on Microblogging by Integrating Text and Image Features

Most studies about sentiment analysis on microblogging usually focus on the features mining from the text. This paper presents a new sentiment analysis method by combing features from text with features from image. Bigram model is applied in text feature extraction while color and texture information are extracted from images. Considering the sentiment classification, we propose a new neighborhood classier based on the similarity of two instances described by the fusion of text and features. Experimental results show that our proposed method can improve the performance significantly on Sina Weibo data (we collect and label the data). We find that our method can not only increasingly improve the F values of the classification comparing with only used text or images features, but also outperforms the NaiveBayes and SVM classifiers using all features with text and images.

Lattice-valued interval sets and t-representable interval set t-norms

Interval sets were proposed by Yiyu Yao in 1993. Interval set theory is related to interval analysis, rough set, fuzzy set and flou set, it can be regarded as a new tool for representing incomplete and uncertain information. In this paper, we introduce the notion of lattice-valued interval set, and investigate its basic algebra operations. We divide the operations into two kinds: inducible and non-inducible. Moreover, we propose the notion t-representable interval set t-norms, and prove a fundamental theorem related residuated implication of t-representable interval set t-norms.

A new extension model of rough sets under incomplete information

The classical rough set theory based on complete information systems stems from the observation that objects with the same characteristics are indiscernible according to available information. With respect to upper-approximation and lower-approximation defined on an indiscernibility relation it classifies objects into different equivalent classes. But in some cases such a rigid indiscernibility relation is far from applications in the real world. Therefore, several generalizations of the rough set theory have been proposed some of which extend the indiscernibility relation using more general similarity or tolerance relations. For example, Kryszkiewicz [4] studied a tolerance relation, and Stefanowski [7] explored a non-symmetric, similarity relation and valued tolerance relation. Unfortunately, All the extensions mentioned above have their inherent limitations. In this paper, after discussing several extension models based on rough sets for incomplete information, a concept of constrained dissymmetrical similarity relation is introduced as a new extension of the rough set theory, the upper-approximation and the lower-approximation defined on constrained similarity relation are proposed as well. Furthermore, we present the comparison between the performance of these extended relations. Analysis of results shows that this relation works effectively in incomplete information and generates rational object classification

Three-Way Decisions Versus Two-Way Decisions on Filtering Spam Email

A three-way decisions solution and a two-way decisions solution for filtering spam emails are examined in this paper. Compared to two-way decisions, the spam filtering is no longer viewed as a binary classification problem, and each incoming email is accepted as a legitimate or rejected as a spam or undecided as a further-examined email in the three-way decisions. One advantage of the three-way decisions solution for spam filtering is that it can reduce the error rate of classifying a legitimate email to spam with minimum misclassification cost. The other one is that the solution can provide a more meaningful decision procedure for users while it is not restricted to a specific classifier. Experimental results on several corpus show that the three-way decisions solution can get a lower error rate and a lower misclassification cost.

An Incremental Updating Algorithm for Core Computing in Dominance-Based Rough Set Model

This paper analyzes incremental updating for core computing in a dominance-based rough set model, which extends previous reduct studies in capability of dynamic updating and dominance relation. Then we redefine the dominance discernibility matrix and present an incremental updating algorithm. In this algorithm, when new samples arrive, the proposed solution only involves a few modifications to relevant rows and columns in the dominance discernibility matrix instead of recalculation.Both of theoretical analysis and experimental results show that the algorithm is effective and efficient in dynamic computation.

Granular-based partial periodic pattern discovery over time series data

Periodicity analysis of the time series is getting more and more significant. There are many contributions for periodic pattern discovery, however, few laid emphasis on the further usage. In the paper, we propose a granular-based partial periodic pattern detecting method over time series data. This method can detect all patterns of every possible periodicity without any prior knowledge of the data sets, by setting different granularity and minimum support threshold. The results that it learned can be used in outlier or change point detection in time series data analysis. The experiment results show its effectiveness.

Selection and optimization of cut-points for numeric attribute values

Data discretization is the process of setting several cut-points which can represent attribute values using different symbols or integer values for continuous numeric attribute values. A hybrid method based on neural network and genetic algorithm is proposed to select and optimize the cut-points for numeric attribute values. The values of cuts are trained through the four-layer neural network and the number of cut-points is optimized by the genetic algorithm. The results for intervals through the presented method can be more precise. The experimental results show that the cut-points are well obtained compared with the other method.

An Incremental Rule Induction Algorithm Based on Ordering Relations

Based on multi-dominance discernibility matrices, an incremental algorithm INRIDDM is proposed by means of dominance-based rough set approach. For the incremental algorithm, when a new object arrives, after updating one row or one column in the matrix, we could get the updated rule sets. Computation analysis and experimental results show that the incremental algorithm INRIDDM is superior to other non-incremental algorithms on dealing with large data sets.

ROGAND: A Discretization Model

A discretization model called ROGAND ( ROugh sets, Genetic Algorithm and Neural network based Discretization ) is presented in the paper, which combines Rough set theory with the genetic algorithm to build a four-layer neural network. This model consists of the data preprocessor (DP), the discretization module(DM) and the optimization module (OM). The discretized intervals obtained through the ROGAND model is independent on the candidates of cut-point sets and the denoted values can be more precise. The experiments indicate that the method is effective and the output cut-points are accurate and easy-set.