Nittaya Kerdprasop

Weighted k-means for density-biased clustering

Clustering is a task of grouping data based on similarity. A popular k-means algorithm groups data by firstly assigning all data points to the closest clusters, then determining the cluster means. The algorithm repeats these two steps until it has converged. We propose a variation called weighted k-means to improve the clustering scalability. To speed up the clustering process, we develop the reservoir-biased sampling as an efficient data reduction technique since it performs a single scan over a data set. Our algorithm has been designed to group data of mixture models. We present an experimental evaluation of the proposed method.

Semantic knowledge integration to support inductive query optimization

We study query evaluation within a framework of inductive databases. An inductive database is a concept of the next generation database in that the repository should contain not only persistent and derived data, but also the patterns of stored data in a unified format. Hence, the database management system should support both data processing and data mining tasks. Having provided with a tightly-coupling environment, users can then interact with the system to create, access, and modify data as well as to induce and query mining patterns. In this paper, we present a framework and techniques of query evaluation in such an environment so that the induced patterns can play a key role as semantic knowledge in the query rewriting and optimization process. Our knowledge induction approach is based on rough set theory. We present the knowledge induction algorithm driven by a users query and explain the method through running examples. The advantages of the proposed techniques are confirmed with experimental results.

Density-biased clustering based on reservoir sampling

Clustering is a task of grouping data based on similarity. A popular k-means algorithm groups data by firstly assigning all data points to the closest clusters, then determining the cluster means. The algorithm repeats these two steps until it has converged. We propose a variation called weighted k-means to improve the clustering scalability. To speed up the clustering process, we develop the reservoir-biased sampling as an efficient data reduction technique since it performs a single scan over a data set. Our algorithm has been designed to group data of mixture models. We present an experimental evaluation of the proposed method.

Data Mining in Semantic Web Data

These specific formats cannot be used directly in most data mining tools. We thus propose a methodology to mine data that appear in an RDF format. The mining process has been demonstrated through the use of R packages.

On Transforming the ER Model to Ontology Using Proté gé OWL Tool

Ontology is a concept to organize domains that can be widely used in many fields. For building the OWL ontology, several existing data sources such as XML, relational databases were used. Most researchers try to map data in a format of relational database into the OWL ontology using OWL syntax, which sometimes is difficult, especially for a person who does not know this syntax, or uses mistaken work to create the OWL ontology. So, in this research, we propose a mechanism to construct OWL ontology in order to reduce the problem of lack of understanding about the OWL ontology syntax based on Entity Relationship Model (ER), which is a model for describing data in a conceptual level of database design. We demonstrate a step-by-step transformation of ER model into OWL ontology using a tool editor called Prote ge . The evaluations of the building ontology will use FaCT++ and HermiT 1.3.8. The results have shown that the ability to convert each part of ER model is very accurate, fast and easy to use.

Predicting Rare Classes of Primary Tumors with Over-Sampling Techniques

The discovery of hidden biomedical patterns from large clinical databases can uncover knowledge to support prognosis and diagnosis decision makings. Researchers and healthcare professionals have applied data mining technology to obtain descriptive patterns and predictive models from biomedical and healthcare databases. However, clinical application of data mining algorithms has a severe problem of low predictive accuracy that hamper their wide usage in the clinical environment. We thus focus our study on the improvement of predictive accuracy of the models created from the data mining algorithms. Our main research interest concerns the problem of learning a tree-based classifier model from a multiclass data set with low prevalence rate of some minority classes. We apply random over-sampling and synthetic minority over-sampling (SMOTE) techniques to increase the predictive performance of the learned model. In our study, we consider specific kinds of primary tumors occurring at the frequency rate less than one percent as rare classes. From the experimental results, the SMOTE technique gave a high specificity model, whereas the random over-sampling produced a high sensitivity classifier. The precision performance of a tree-based model obtained from the random over-sampling technique is on average much better than the model learned from the original imbalanced data set.

Density Estimation Technique for Data Stream Classification

Density estimation is an important pre-processing step in the problem of data stream classification in which the number of data is overwhelming and the exact data distribution is unknown. We simplify the problem by employing a statistical sampling technique to obtain an approximate solution. With the proposed method, an unbounded large data set can be sampled in a number of random configurations, and that data can be used to describe the data set as a whole. The efficiency of the method depends largely on the ability to draw samples effectively which in turn depends on how close we can estimate the target density. We use finite mixture models to represent the probability density functions of the data stream. Then, we apply the EM algorithm twice to learn the model parameters. The efficiency of our estimation technique has been shown in the experimental results

Integrating Inductive Knowledge into the Inference System of Biomedical Informatics

This paper presents a new methodology for the design and implementation of the next generation rule-based expert system in a medical domain. In addition to the set of predefined rules, the system includes rules that are automatically induced from the database instances. We design the inductive expert system such that the inductive process has been done through the tree-based knowledge discovery technique. Probabilistic decision rules are then transformed from the induced decision tree. The induced, as well as predefined, rules together form a knowledge base for the inductive expert system. Another feature of our system is the inference engine that can be created automatically. The system is intended to support decision making in biomedical informatics. The general design of our system is, however, appropriate for other domains as well.

Computational Intelligence Techniques to Fault Detection in the Semiconductor Manufacturing Process

The semiconductor industry deals with the production at a scale of nanometer, thus resulting in the process control with little margin of error. Timely detection of faults during the manufacturing process is critical to the improvement in product yields. Difficulty of detecting accurately faulty processes and products is due to the abundant of data obtained from hundreds of tool-state and process-state sensors. We thus analyze this problem through the computational intelligence techniques. The analysis results reveal the minimal set of features for fault detection as well as the high precision classification model of faults.

SUT-Miner: A Knowledge Mining and Managing System for Medical Databases

Knowledge is a valuable asset to most organizations as a substantial source to support better decisions. Recently there has been an increasing interest in devising database and data mining technologies to automatically induce knowledge from biomedicine, clinical and health data. Most work had adopted a single technique in the knowledge induction process. We propose a knowledge mining system as an integrated environment storing a repertoire of tools for discovering strong and useful knowledge. We demonstrate the usefulness aspect through the semi-automatic trigger creation for the medical database. A rapid prototyping of association mining engine is also presented in the paper.