Einoshin Suzuki

Exception Rule Mining with a Relative Interestingness Measure

This paper presents a method for mining exception rules based on a novel measure which estimates interestingness relative to its corresponding common sense rule and reference rule. Mining interesting rules is one of the important data mining tasks. Interesting rules bring novel knowledge that helps decision makers for advantageous actions. It is true that interestingness is a relative issue that depends on the other prior knowledge. However, this estimation can be biased due to the incomplete or inaccurate knowledge about the domain. Even if possible to estimate interestingness, it is not so trivial to judge the interestingness from a huge set of mined rules. Therefore, an automated system is required that can exploit the knowledge extractacted from the data in measuring interestingness. Since the extracted knowledge comes from the data, so it is possible to find a measure that is unbiased from the users own belief. An unbiased measure that can estimate the interestingness of a rule with respect to the extractacted rules can be more acceptable to the user. In this work we try to show through the experiments, how our proposed relative measure can give an unbiased estimate of relative interestingness in a rule considering already mined rules.

Hierarchical Gaussian Descriptor for Person Re-identification

Describing the color and textural information of a person image is one of the most crucial aspects of person re-identification. In this paper, we present a novel descriptor based on a hierarchical distribution of pixel features. A hierarchical covariance descriptor has been successfully applied for image classification. However, the mean information of pixel features, which is absent in covariance, tends to be major discriminative information of person images. To solve this problem, we describe a local region in an image via hierarchical Gaussian distribution in which both means and covariances are included in their parameters. More specifically, we model the region as a set of multiple Gaussian distributions in which each Gaussian represents the appearance of a local patch. The characteristics of the set of Gaussians are again described by another Gaussian distribution. In both steps, unlike the hierarchical covariance descriptor, the proposed descriptor can model both the mean and the covariance information of pixel features properly. The results of experiments conducted on five databases indicate that the proposed descriptor exhibits remarkably high performance which outperforms the state-of-the-art descriptors for person re-identification.

Discovery of Surprising Exception Rules Based on Intensity of Implication

This paper presents an algorithm for discovering surprising exception rules from data sets. An exception rule, which is defined as a deviational pattern to a common sense, exhibits unexpectedness and is sometimes extremely useful. A domain-independent approach, PEDRE, exists for the simultaneous discovery of exception rules and their common sense rules. However, PEDRE, being too conservative, have difficulty in discovering surprising rules. Historic exception discoveries show that surprise is often linked with interestingness. In order to formalize this notion we propose a novel approach by improving PEDRE. First, we reformalize the problem and settle a looser constraints on the reliability of an exception rule. Then, in order to screen out uninteresting rules, we introduce, for an exception rule, an evaluation criterion of surprise by modifying intensity of implication, which is based on significance. Our approach has been validated using data sets from the UCI repository.

UNDIRECTED DISCOVERY OF INTERESTING EXCEPTION RULES

This paper presents an efficient algorithm for discovering exception rules from a data set without domain-specific information. An exception rule, which is defined as a deviational pattern to a strong rule, exhibits unexpectedness and is sometimes extremely useful. Previous discovery approaches for this type of knowledge can be classified into a directed approach, which obtains exception rules each of which deviates from a set of user-prespecified strong rules, and an undirected approach, which typically discovers a set of rule pairs each of which represents a pair of an exception rule and its corresponding strong rule. It has been pointed out that unexpectedness is often related to interestingness. In this sense, an undirected approach is promising since its discovery outcome is free from human prejudice and thus tends to be highly unexpected. However, this approach is prohibitive due to extra search for strong rules as well as unreliable patterns in the output. In order to circumvent these difficulties we propose a method based on sound pruning and probabilistic estimation. The sound pruning reduces search time to a reasonable amount, and enables exhaustive search for rule pairs. The normal approximations of the multinomial distributions are employed as the method for evaluating reliability of a rule pair. Our method has been validated using two medical data sets under supervision of a physician and two benchmark data sets in the machine learning community.

Unified algorithm for undirected discovery of exception rules

This article presents an algorithm that seeks every possible exception rule that violates a commonsense rule and satisfies several assumptions of simplicity. Exception rules, which represent systematic deviation from commonsense rules, are often found interesting. Discovery of pairs that consist of a commonsense rule and an exception rule, resulting from undirected search for unexpected exception rules, was successful in various domains. In the past, however, an exception rule represented a change of conclusion caused by adding an extra condition to the premise of a commonsense rule. That approach formalized only one type of exception and failed to represent other types. To provide a systematic treatment of exceptions, we categorize exception rules into 11 categories, and we propose a unified algorithm for discovering all of them. Preliminary results on 15 real‐world datasets provide an empirical proof of effectiveness of our algorithm in discovering interesting knowledge. The empirical results also match our theoretical analysis of exceptions, showing that the 11 types can be partitioned in three classes according to the frequency with which they occur in data.

Discovering community-oriented roles of nodes in a social network

We propose a new method for identifying the role of a vertex in a social network. Existing well-known metrics of node centrality such as betweenness, degree and closeness do not take the community structure within a network into consideration. Furthermore, existing proposed community-based roles are defined using cliques, and thereby it is difficult to discover vertices with only few links that bridge communities. To overcome the shortcomings, we propose three community-oriented roles, bridges, gateways and hubs, without knowledge on the community structure, for representing vertices that bridge communities. We believe that detecting the roles in a social network is useful because such nodes are valuable by themselves due to their intermediate roles between communities and also because the nodes are likely to provide a deeper understanding of the communities. Our method outperforms the state-of-the-art method through experiments using data of DBLP records in terms of the subjective validness of the outputs.

Bloomy Decision Tree for Multi-objective Classification

This paper presents a novel decision-tree induction for a multi-objective data set, i.e. a data set with a multi-dimensional class. Inductive decision-tree learning is one of the frequently-used methods for a single-objective data set, i.e. a data set with a single-dimensional class. However, in a real data analysis, we usually have multiple objectives, and a classifier which explains them simultaneously would be useful and would exhibit higher readability. A conventional decision-tree inducer requires transformation of a multi-dimensional class into a single-dimensional class, but such a transformation can considerably worsen both accuracy and readability. In order to circumvent this problem we propose a bloomy decision tree which deals with a multi-dimensional class without such transformations. A bloomy decision tree has a set of split nodes each of which splits examples according to their attribute values, and a set of flower nodes each of which predicts a class dimension of examples. A flower node appears not only at the fringe of a tree but also inside a tree. Our pruning is executed during tree construction, and evaluates each class dimension based on CramErs V. The proposed method has been implemented as D3-B (Decision tree in Bloom), and tested with eleven data sets. The experiments showed that D3-B has higher accuracies in nine data sets than C4.5 and tied with it in the other two data sets. In terms of readability, D3-B has a smaller number of split nodes in all data sets, and thus outperforms C4.5.

Scheduled Discovery of Exception Rules

This paper presents an algorithm for discovering pairs of an exception rule and a common sense rule under a prespecified schedule. An exception rule, which represents a regularity of exceptions to a common sense rule, often exhibits interestingness. Discovery of pairs of an exception rule and a common sense rule has been successful in various domains. In this method, however, both the number of discovered rules and time needed for discovery depend on the values of thresholds, and an appropriate choice of them requires expertise on the data set and on the discovery algorithm. In order to circumvent this problem, we propose two scheduling policies for updating values of these thresholds based on a novel data structure. The data structure consists of multiple balanced search-trees, and efficiently manages discovered patterns with multiple indices. One of the policies represents a full specification of updating by the user, and the other iteratively improves a threshold value by deleting the worst pattern with respect to its corresponding index. Preliminary results on four real-world data sets are highly promising. Our algorithm settled values of thresholds appropriately, and discovered interesting exception-rules from all these data sets.

Detecting interesting exceptions from medical test data with visual summarization

We propose a method which visualizes irregular multidimensional time-series data as a sequence of probabilistic prototypes for detecting exceptions from medical test data. Conventional visualization methods often require iterative analysis and considerable skill thus are not totally supported by a wide range of medical experts. Our PrototypeLines displays summarized information based on a probabilistic mixture model by using hue only thus is considered to exhibit novelty. The effectiveness of the summarization is pursued mainly through use of a novel information criterion. We report our endeavor with chronic hepatitis data, especially discoveries of interesting exceptions by a nonexpert and an untrained expert.

Evaluating Hypothesis-Driven Exception-Rule Discovery with Medical Data Sets

This paper presents a validation, with two common medical data sets, of exception-rule discovery based on a hypothesis-driven approach. The analysis confirmed the effectiveness of the approach in discovering valid, novel and surprising knowledge.