AI 2010 Advances in Artificial Intelligence %2D 23rd Australasian Joint Conference

On the Effectiveness of Gene Selection for Microarray Classification Methods.

Privacy Protection for Genomic Data Current Techniques and Challenges.

Satisfying Privacy Requirements One Step before Anonymization.

Privacy Protection for Genomic Data Current Techniques and Challenges.

Identifying functional miRNA%2DmRNA regulatory modules with correspondence latent dirichlet allocation.

Kernel Discriminant Learning for Ordinal Regression.

Feature fusion using locally linear embedding for classification.

Microdata Protection Through Approximate Microaggregation.
Microdata protection is a hot topic in the field of Statistical Disclosure Control, which has gained special interest after the disclosure of 658000 queries by the America Online (AOL) search engine in August 2006. Many algorithms, methods and properties have been proposed to deal with microdata disclosure. One of the emerging concepts in microdata protection is k-anonymity, introduced by Samarati and Sweeney. k-anonymity provides a simple and efficient approach to protect private individual information and is gaining increasing popularity. k-anonymity requires that every record in the microdata table released be indistinguishably related to no fewer than k respondents. In this paper, we apply the concept of entropy to propose a distance metric to evaluate the amount of mutual information among records in microdata, and propose a method of constructing dependency tree to find the key attributes, which we then use to process approximate microaggregation. Further, we adopt this new microaggregation technique to study k-anonymity problem, and an efficient algorithm is developed. Experimental results show that the proposed microaggregation technique is efficient and effective in the terms of running time and information loss. Keywords: Microdata Protection, Privacy; 1
Finding Irredundant Contained Rewritings of Tree Pattern Queries Using Views.

Injecting purpose and trust into data anonymisation.

Achieving P%2DSensitive K%2DAnonymity via Anatomy.

Efficient discovery of risk patterns in medical data.
Objective: This paper studies a problem of efficiently discovering risk patterns in medical data. Risk patterns are defined by a statistical metric, relative risk, which has been widely used in epidemiological research. Methods: To avoid fruitless search in the complete exploration of risk patterns, we define optimal risk pattern set to exclude superfluous patterns, i.e. complicated patterns with lower relative risk than their corresponding simpler form patterns. We prove that mining optimal risk pattern sets conforms an anti-monotone property that supports an efficient mining algorithm. We propose an efficient algorithm for mining optimal risk pattern sets based on this property. We also propose a hierarchical structure to present discovered patterns for the easy perusal by domain experts. Results: The proposed approach is compared with two well-known rule discovery methods, decision tree and association rule mining approaches on benchmark data sets and applied to a real world application. The proposed method discovers more and better quality risk patterns than a decision tree approach. The decision tree method is not designed for such applications and is inadequate for pattern exploring. The
Exploring complex miRNA%2DmRNA interactions with Bayesian networks by splitting%2Daveraging strategy.

An integrated model for next page access prediction.

%28alpha%2C k%29%2Danonymous data publishing.
E-mail address:
Discovery of functional miRNA%2DmRNA regulatory modules with computational methods.

Data Mining and Analytics 2008

Integrating recommendation models for improved web page prediction accuracy.

On the Complexity of Restricted k%2Danonymity Problem.

Priority Driven K%2DAnonymisation for Privacy Protection.
Given the threat of re-identification in our growing digital society, guaranteeing privacy while providing worthwhile data for knowledge discovery has become a difficult problem. k-anonymity is a major technique used to ensure privacy by generalizing and suppressing attributes and has been the focus of intense research in the last few years. However, data modification techniques like generalization may produce anonymous data unusable for medical studies because some attributes become too coarse-grained. In this paper, we propose a priority driven k-anonymisation that allows to specify the degree of acceptable distortion for each attribute separately. We also define some appropriate metrics to measure the distance and information loss, which are suitable for both numerical and categorical attributes. Further, we formulate the priority driven k-anonymisation as the k-nearest neighbor (KNN) clustering problem by adding a constraint that each cluster contains at least k tuples. We develop an efficient algorithm for priority driven k-anonymisation. Experimental results show that the proposed technique causes significantly less distortions. Keywords: K-Anonymity; Privacy Protection; 1
L%2DDiversity Based Dynamic Update for Large Time%2DEvolving Microdata.

Portable devices of security and privacy preservation for e%2Dlearning.

Authorization approaches for advanced permission%2Drole assignments.

A Role%2DBased Framework for Multi%2Dagent Teaming.

Anonymization by Local Recoding in Data with Attribute Hierarchical Taxonomies.
Abstract—Individual privacy will be at risk if a published data set is not properly deidentified. k-Anonymity is a major technique to deidentify a data set. Among a number of k-anonymization schemes, local recoding methods are promising for minimizing the distortion of a k-anonymity view. This paper addresses two major issues in local recoding k-anonymization in attribute hierarchical taxonomies. First, we define a proper distance metric to achieve local recoding generalization with small distortion. Second, we propose a means to control the inconsistency of attribute domains in a generalized view by local recoding. We show experimentally that our proposed local recoding method based on the proposed distance metric produces higher quality k-anonymity tables in three quality measures than a global recoding anonymization method, Incognito, and a multidimensional recoding anonymization method, Multi. The proposed inconsistency handling method is able to balance distortion and consistency of a generalized view. Index Terms—k-anonymization, local recoding, generalization distance, inconsistency. Ç
A role%2Doriented BDI framework for real%2Dtime multiagent teaming.

Privacy preserving serial data publishing by role composition.
Previous works about privacy preserving serial data publishing on dynamic databases have relied on unrealistic assumptions of the nature of dynamic databases. In many applications, some sensitive values changes freely while others never change. For example, in medical applications, the disease attribute changes with time when patients recover from one disease and develop another disease. However, patients do not recover from some diseases such as HIV. We call such diseases permanent sensitive values. To the best of our knowledge, none of the existing solutions handle these realistic issues. We propose a novel anonymization approach called HD-composition to solve the above problems. Extensive experiments with real data confirm our theoretical results. 1.
Enhanced P%2DSensitive K%2DAnonymity Models for Privacy Preserving Data Publishing.
Abstract. Publishing data for analysis from a micro data table containing sensitive attributes, while maintaining individual privacy, is a problem of increasing significance today. The k-anonymity model was proposed for privacy preserving data publication. While focusing on identity disclosure, k-anonymity model fails to protect attribute disclosure to some extent. Many efforts are made to enhance the k-anonymity model recently. In this paper, we propose two new privacy protection models called (p, α)-sensitive k-anonymity and (p +, α)-sensitive k-anonymity, respectively. Different from previous the p-sensitive k-anonymity model, these new introduced models allow us to release a lot more information without compromising privacy. Moreover, we prove that the (p, α)-sensitive and (p +, α)-sensitive k-anonymity problems are NP-hard. We also include testing and heuristic generating algorithms to generate desired micro data table. Experimental results show that our introduced model could significantly reduce the privacy breach. 1
Data Mining and Analytics 2007



Data Mining and Analytics 2006
by the CRC for Construction Innovation. This delegation is contained in Clause 30
A framework for role%2Dbased group deligation in distributed environments.

A Framework of Combining Markov Model With Association Rules for Predicting Web Page Accesses.
The importance of predicting Web users ’ behaviour and their next movement has been recognised and discussed by many researchers lately. Association rules and Markov models are the most commonly used approaches for this type of prediction. Association rules tend to generate many rules, which result in contradictory predictions for a user session. Low order Markov models do not use enough user browsing history and therefore, lack accuracy, whereas, high order Markov models incur high state space complexity. This paper proposes a novel approach that integrates both association rules and low order Markov models in order to achieve higher accuracy with low state space complexity. A low order Markov model provides high coverage with low state space complexity, and association rules help achieve better accuracy.
A Comparative Study of Classification Methods For Microarray Data Analysis.
In response to the rapid development of DNA Microarray technology, many classification methods have been used for Microarray classification. SVMs, decision trees, Bagging, Boosting and Random Forest are commonly used methods. In this paper, we conduct experimental comparison of LibSVMs, C4.5, BaggingC4.5, AdaBoostingC4.5, and Random Forest on seven Microarray cancer data sets. The experimental results show that all ensemble methods outperform C4.5. The experimental results also show that all five methods benefit from data preprocessing, including gene selection and discretization, in classification accuracy. In addition to comparing the average accuracies of ten-fold cross validation tests on seven data sets, we use two statistical tests to validate findings. We observe that Wilcoxon signed rank test is better than sign test for such purpose.
Analysis of Breast Feeding Data Using Data Mining Methods.
The purpose of this study is to demonstrate the benefit of using common data mining techniques on survey data where statistical analysis is routinely applied. The statistical survey is commonly used to collect quantitative information about an item in a population. Statistical analysis is usually carried out on survey data to test hypothesis. We report in this paper an application of data mining methodologies to breast feeding survey data which have been conducted and analysed by statisticians. The purpose of the research is to study the factors leading to deciding whether or not to breast feed a new born baby. Various data mining methods are applied to the data. Feature or variable selection is conducted to select the most discriminative and least redundant features using an information theory based method and a statistical approach. Decision tree and regression approaches are tested on classification tasks using features selected. Risk pattern mining method is also applied to identify groups with high risk of not breast feeding. The success of data mining in this study suggests that using data mining approaches will be applicable to other similar survey data. The data mining methods, which enable a search for hypotheses, may be used as a complementary survey data analysis tool to traditional statistical analysis.
Analysis of Breast Feeding Data Using Data Mining Methods.
The purpose of this study is to demonstrate the benefit of using common data mining techniques on survey data where statistical analysis is routinely applied. The statistical survey is commonly used to collect quantitative information about an item in a population. Statistical analysis is usually carried out on survey data to test hypothesis. We report in this paper an application of data mining methodologies to breast feeding survey data which have been conducted and analysed by statisticians. The purpose of the research is to study the factors leading to deciding whether or not to breast feed a new born baby. Various data mining methods are applied to the data. Feature or variable selection is conducted to select the most discriminative and least redundant features using an information theory based method and a statistical approach. Decision tree and regression approaches are tested on classification tasks using features selected. Risk pattern mining method is also applied to identify groups with high risk of not breast feeding. The success of data mining in this study suggests that using data mining approaches will be applicable to other similar survey data. The data mining methods, which enable a search for hypotheses, may be used as a complementary survey data analysis tool to traditional statistical analysis.
Achieving k%2DAnonymity by Clustering in Attribute Hierarchical Structures.
Abstract. Individual privacy will be at risk if a published data set is not properly deidentified. k-anonymity is a major technique to de-identify a data set. A more general view of k-anonymity is clustering with a constraint of the minimum number of objects in every cluster. Most existing approaches to achieving k-anonymity by clustering are for numerical (or ordinal) attributes. In this paper, we study achieving k-anonymity by clustering in attribute hierarchical structures. We define generalisation distances between tuples to characterise distortions by generalisations and discuss the properties of the distances. We conclude that the generalisation distance is a metric distance. We propose an efficient clustering-based algorithm for k-anonymisation. We experimentally show that the proposed method is more scalable and causes significantly less distortions than an optimal global recoding k-anonymity method. 1
%28alpha
. Frequently, data in scientific computing is in its abstract form a finite point set in space, and it is sometimes useful or required to compute what one might call the &#034;shape&#034; of the set. For that purpose, this paper introduces the formal notion of the family of ff-shapes of a finite point set in IR  3  . Each shape is a well-defined polytope, derived from the Delaunay triangulation of the point set, with a parameter ff 2 IR controlling the desired level of detail. An algorithm is presented that constructs the entire family of shapes for a given set of size n in time O(n  2  ), worst case. A robust implementation of the algorithm is discussed and several applications in the area of scientific computing are mentioned. Categories and Subject Descriptors. F.2.2 [Analysis of Algorithms and Problem Complexity ]: Nonnumerical algorithms and problems --- geometrical problems and computations; G.4 [Mathe-  matics of Computing]: Mathematical Software --- reliability and robustness; I.2.10 [Ar...
Classification Using Multiple and Negative Target Rules.

Combined Gene Selection Methods for Microarray Data Analysis.

On Optimal Rule Discovery.
Abstract—In machine learning and data mining, heuristic and association rules are two dominant schemes for rule discovery. Heuristic rule discovery usually produces a small set of accurate rules, but fails to find many globally optimal rules. Association rule discovery generates all rules satisfying some constraints, but yields too many rules and is infeasible when the minimum support is small. Here, we present a unified framework for the discovery of a family of optimal rule sets and characterize the relationships with other rule-discovery schemes such as nonredundant association rule discovery. We theoretically and empirically show that optimal rule discovery is significantly more efficient than association rule discovery independent of data structure and implementation. Optimal rule discovery is an efficient alternative to association rule discovery, especially when the minimum support is low. Index Terms—Data mining, rule discovery, optimal rule set. 1
Robust Rule%2DBased Prediction.
Abstract—This paper studies a problem of robust rule-based classification, i.e., making predictions in the presence of missing values in data. This study differs from other missing value handling research in that it does not handle missing values but builds a rule-based classification model to tolerate missing values. Based on a commonly used rule-based classification model, we characterize the robustness of a hierarchy of rule sets as k-optimal rule sets with the decreasing size corresponding to the decreasing robustness. We build classifiers based on k-optimal rule sets and show experimentally that they are more robust than some benchmark rule-based classifiers, such as C4.5rules and CBA. We also show that the proposed approach is better than two well-known missing value handling methods for missing values in test data. Index Terms—Data mining, rule, classification, robustness. 1
Using multiple and negative target rules to make classifiers more understandable.

Using Association Rules to Make Rule%2Dbased Classifiers Robust.
Rule-based classification systems have been widely used in real world applications because of the easy interpretability of rules. Many traditional rule-based classifiers prefer small rule sets to large rule sets, but small classifiers are sensitive to the missing values in unseen test data. In this paper, we present a larger classifier that is less sensitive to the missing values in unseen test data. We experimentally show that it is more accurate than some benchmark classifies when unseen test data have missing values.
Using Association Rules to Make Rule%2Dbased Classifiers Robust.
Rule-based classification systems have been widely used in real world applications because of the easy interpretability of rules. Many traditional rule-based classifiers prefer small rule sets to large rule sets, but small classifiers are sensitive to the missing values in unseen test data. In this paper, we present a larger classifier that is less sensitive to the missing values in unseen test data. We experimentally show that it is more accurate than some benchmark classifies when unseen test data have missing values.
Finding Similar Patterns in Microarray Data.

Mining risk patterns in medical data.
In this paper, we discuss a problem of finding risk patterns in medical data. We define risk patterns by a statistical metric, relative risk, which has been widely used in epidemiological research. We characterise the problem of mining risk patterns as an optimal rule discovery problem. We study an anti-monotone property for mining optimal risk pattern sets and present an algorithm to make use of the property in risk pattern discovery. The method has been applied to a real world data set to find patterns associated with an allergic event for ACE inhibitors. The algorithm has generated some useful results for medical researchers. Categories and Subject Descriptors
Representing Association Classification Rules Mined from Health Data.
Abstract. An association classification algorithm has been developed to explore adverse drug reactions in a large medical transaction dataset with unbalanced classes. Rules discovered can be used to alert medical practitioners when prescribing drugs, to certain categories of patients, to potential adverse effects. We assess the rules using survival charts and propose two kinds of probability trees to present them. Both of them represent the risk of given adverse drug reaction for certain categories of patients in terms of risk ratios, which are familiar to medical practitioners. The first approach shows risk ratios when all rule conditions apply. The second presents the risk associated with a single risk factor with other parts of the rule identifying the cohort of the patient subpopulation. Thus, the probability trees can present clearly the risk of specific adverse drug reactions to prescribers. 1
Mining Informative Rule Set for Prediction.

Association Rule Discovery with Unbalanced Class Distributions.
There are many methods for finding association rules in very large data. However it is well known that most general association rule discovery methods find too many rules, which include a lot of uninteresting rules. Furthermore, the performances of many such algorithms deteriorate when the minimum support is low. They fail to find many interesting rules even when support is low, particularly in the case of significantly unbalanced classes. In this paper we present an algorithm which finds association rules based on a set of new interestingness criteria. The algorithm is applied to a real-world health data set and successfully identifies groups of patients with high risk of adverse reaction to certain drugs. A statistically guided method of selecting appropriate features has also been developed. Initial results have shown that the proposed algorithm can find interesting patterns from data sets with unbalanced class distributions without performance loss.
Direct Interesting Rule Generation.
An association rule generation algorithm usually generates too many rules including a lot of uninteresting ones. Many interestingness criteria are proposed to prune those uninteresting rules. However, they work in post-pruning process and hence do not improve the rule generation ef-£ciency. In this paper, we discuss properties of informative rule set and conclude that the informative rule set includes all interesting rules measured by many commonly used interestingness criteria, and that rules excluded by the informative rule set are forwardly prunable, i.e. they can be removed in the rule generation process instead of post pruning. Based on these properties, we propose a Direct Interesting rule Generation algorithm, DIG, to directly generate interesting rules de£ned by any of 12 interestingness criteria discussed in this paper. We further show experimentally that DIG is faster and uses less memory than Apriori. 
Construct robust rule sets for classification.
We study the problem of computing classification rule sets from relational databases so that accurate predictions can be made on test data with missing attribute values. Traditional classifiers perform badly when test data are not as complete as the training data because they tailor a training database too much. We introduce the concept of one rule set being more robust than another, that is, able to make more accurate predictions on test data with missing attribute values. We show that the optimal class association rule set is as robust as the complete class association rule set. We then introduce the k-optimal rule set, which provides predictions exactly the same as the optimal class association rule set on test data with up to k missing attribute values. This leads to a hierarchy of k-optimal rule sets in which decreasing size corresponds to decreasing robustness, and they all more robust than a traditional classification rule set. We introduce two methods to find k-optimal rule sets, i.e. an optimal association rule mining approach and a heuristic approximate approach. We show experimentally that a k-optimal rule set generated by the optimal association rule mining approach performs better than that by the heuristic approximate approach and both rule sets perform significantly better than a typical classification rule set (C4.5Rules) on incomplete test data.
Mining the optimal class association rule set.
Rule-based classification systems have been widely used in real world applications because of the easy interpretability of rules. Many traditional rule-based classifiers prefer small rule sets to large rule sets, but small classifiers are sensitive to the missing values in unseen test data. In this paper, we present a larger classifier that is less sensitive to the missing values in unseen test data. We experimentally show that it is more accurate than some benchmark classifies when unseen test data have missing values.

Mining the Smallest Association Rule Set for Predictions.
Mining transaction databases for association rules usually generates a large number of rules, most of which are unnecessary when used for subsequent prediction. In this paper we define a rule set for a given transaction database that is much smaller than the association rule set but makes the same predictions as the association rule set by the confidence priority. We call this subset the informative rule set. The informative rule set is not constrained to particular target items; and it is smaller than the non-redundant association rule set. We present an algorithm to directly generate the informative rule set, i.e., without generating all frequent itemsets first, and that accesses the database less often than other unconstrained direct methods. We show experimentally that the informative rule set is much smaller than both the association rule set and the non-redundant association rule set, and that it can be generated more efficiently.

An Adaptive Method of Numerical Attribute Merging for Quantitative Association Rule Mining.

Mining Significant Association Rules %28Short Version%29.

Knowledge Network Based Association Discovery %28Preliminary Version%29.

