Miyoung Shin

Empirical data modeling in software engineering using radial basis functions

Many empirical studies in software engineering involve relationships between various process and product characteristics derived via linear regression analysis. We propose an alternative modeling approach using radial basis functions (RBFs) which provide a flexible way to generalize linear regression function. Further, RBF models possess strong mathematical properties of universal and best approximation. We present an objective modeling methodology for determining model parameters using our recent SG algorithm, followed by a model selection procedure based on generalization ability. Finally, we describe a detailed RBF modeling study for software effort estimation using a well-known NASA dataset.

Optimization of subsequence matching under time warping in time-series databases

This paper discusses effective processing of subsequence matching under time warping in time-series databases. Time warping is a transformation that enables finding of sequences with similar patterns even when they are of different lengths. Through a preliminary experiment, we first point out that Naive-Scan, a basic method for processing of subsequence matching under time warping, has its performance bottleneck in the CPU processing step. For optimizing this step, in this paper, we propose a novel method that eliminates all possible redundant calculations. It is verified that this method is not only an optimal one for processing Naive-Scan, but also does not incur any false dismissals. Our experimental results showed that the proposed method can make great improvement in performance of subsequence matching under time warping. Especially, Naive-Scan, which has been known to show the worst performance, performs much better than LB-Scan as well as ST-Filter in all the cases by employing the proposed method for CPU processing. This result is interesting and valuable in that the performance inversion among Naive-Scan, LB-Scan, and ST-Filter has occurred by optimizing the CPU processing step, which is their common performance bottleneck.

Path prediction of moving objects on road networks through analyzing past trajectories

This paper addresses a series of techniques for predicting a future path of an object moving on a road network. Most prior methods for future prediction mainly focus on the objects moving over Euclidean space. A variety of applications such as telematics, however, require us to handle the objects that move over road networks. In this paper, we propose a novel method for predicting a future path of an object in an efficient way by analyzing past trajectories whose changing pattern is similar to that of a current trajectory of a query object. For this purpose, we devise a new function for measuring a similarity between trajectories by considering the characteristics of road networks. By using this function, we search for candidate trajectories whose subtrajectories are similar to a given query trajectory by accessing past trajectories stored in moving object databases. Then, we predict a future path of a query object by analyzing the moving paths along with a current position to a destination of candidate trajectories. Also, we suggest a method that improves the accuracy of path prediction by grouping those moving paths whose differences are not significant.

Gaussian radial basis functions for simulation metamodeling

The paper presents a novel approach for developing simulation metamodels using Gaussian radial basis functions. This approach is based on some recently developed mathematical results for radial basis functions. It is systematic, explicitly controls the underfitting and overfitting tradeoff, and uses a fast computational algorithm that requires minimal human involvement. This approach is illustrated by developing metamodels for the M/M/1 queueing system.

An integrative model of multi-organ drug-induced toxicity prediction using gene-expression data

BackgroundIn practice, some drugs produce a number of negative biological effects that can mitigate their effectiveness as a remedy. To address this issue, several studies have been performed for the prediction of drug-induced toxicity from gene-expression data, and a significant amount of work has been done on predicting limited drug-induced symptoms or single-organ toxicity. Since drugs often lead to some injuries in several organs like liver or kidney, however, it would be very useful to forecast the drug-induced injuries for multiple organs. Therefore, in this work, our aim was to develop a multi-organ toxicity prediction model using an integrative model of gene-expression data.ResultsTo train our integrative model, we used 3708 in-vivo samples of gene-expression profiles exposed to one of 41 drugs related to 21 distinct physiological changes divided between liver and kidney (liver 11, kidney 10). Specifically, we used the gene-expression profiles to learn an ensemble classifier for each of 21 pathology prediction models. Subsequently, these classifiers were combined with weights to generate an integrative model for each pathological finding. The integrative model outputs the likeliness of presenting the trained pathology in a given test sample of gene-expression profile, called an integrative prediction score (IPS). For the evaluation of an integrative model, we estimated the prediction performance with the k-fold cross-validation. Our results demonstrate that the proposed integrative model is superior to individual pathology prediction models in predicting multi-organ drug-induced toxicities over all the targeted pathological findings. On average, the AUC of the integrative models was 88% while the AUC of individual pathology prediction models was 68%.ConclusionsNot only does this integrative model produce comparable prediction performance to existing approaches, but also it produces very stable performance overall. In addition, our approach is easily expandable to a variety of other multi-organ toxicology applications.

Identifying Biologically Significant Pathways by Gene Set Enrichment Analysis Using Fisher's Criterion

Gene set enrichment analysis (GSEA) is a computational method to identify statistically significant gene-sets showing differential expression between two groups. In particular, unlike other previous approaches, this enables us to uncover their biological meanings in an elegant way by providing a unified analytical framework that employs a priori known biological knowledges along with gene expression profiles during the analysis procedure. For original GSEA, all the genes in a given dataset are ordered by the signal-to-noise ratio of their microarray expression profiles between two groups and then further analyses are proceeded. Despite of its impressive results in previous studies with original GSEA, however, gene ranking by the signal-to-noise ratio makes it difficult to extract both highly up-regulated genes and highly down-regulated genes at a time as significant genes, which may not reflect such situations as incurred in metabolic and signaling pathways. Thus, it is necessary to make further investigation for better finding of biologically significant pathways. To deal with this problem, in this article, we explore the method of gene set enrichment analysis with Fishers criterion for gene ranking, named FC-GSEA, and evaluate its effects made in leukemia related pathway analyses.

Parsimonious classifiers for software quality assessment

Modeling to predict fault-proneness of software modules is an important area of research in software engineering. Most such models employ a large number of basic and derived metrics as predictors. This paper presents modeling results based on only two metrics, lines of code and cyclomatic complexity, using radial basis functions with Gaussian kernels as classifiers. Results from two NASA systems are presented and analyzed.

A hybrid approach to gene ranking using gene relation networks derived from literature for the identification of disease gene markers

For the identification of gene markers involved in diseases, microarray expression profiles have been widely used to prioritize genes. In this paper, we propose a novel approach to gene ranking that employs gene relation network derived from literature along with microarray expression scores to calculate ranking statistics of individual genes. In particular, the gene relation network is constructed from literature by applying syntactic analysis and co-occurrence method in a hybrid manner. For evaluation, the proposed method was tested with publicly available prostate cancer data. The result shows that our method is superior to other existing approaches.

A Systematic Framework for Drug Repositioning from Integrated Omics and Drug Phenotype Profiles Using Pathway-Drug Network

Drug repositioning offers new clinical indications for old drugs. Recently, many computational approaches have been developed to repurpose marketed drugs in human diseases by mining various of biological data including disease expression profiles, pathways, drug phenotype expression profiles, and chemical structure data. However, despite encouraging results, a comprehensive and efficient computational drug repositioning approach is needed that includes the high-level integration of available resources. In this study, we propose a systematic framework employing experimental genomic knowledge and pharmaceutical knowledge to reposition drugs for a specific disease. Specifically, we first obtain experimental genomic knowledge from disease gene expression profiles and pharmaceutical knowledge from drug phenotype expression profiles and construct a pathway-drug network representing a priori known associations between drugs and pathways. To discover promising candidates for drug repositioning, we initialize node labels for the pathway-drug network using identified disease pathways and known drugs associated with the phenotype of interest and perform network propagation in a semisupervised manner. To evaluate our method, we conducted some experiments to reposition 1309 drugs based on four different breast cancer datasets and verified the results of promising candidate drugs for breast cancer by a two-step validation procedure. Consequently, our experimental results showed that the proposed framework is quite useful approach to discover promising candidates for breast cancer treatment.

Performance of TPR*-Trees for Predicting Future Positions of Moving Objects in U-Cities

The TPR*-tree is the most widely-used index structure for effectively predicting the future positions of moving objects. The TPR*-tree, however, has the problem that both of the dead spacein a bounding region and the overlap among bounding regions become larger as the prediction time point in the future gets farther. This makes more nodes within the TPR*-tree accessed in query processing time, which incurs serious performance degradation. In this paper, we examine the performance problem quantitatively via a series of experiments. First, we show how much the performance deteriorates as a prediction time point gets farther from the present, and also show how the frequent updates of positions of moving objects alleviate this problem. Our contribution would help provide important clues to devise strategies improving the performance of TPR*-trees further.