Efficient construction of histograms for multidimensional data using quad-trees

Histograms can be useful in estimating the selectivity of queries in areas such as database query optimization and data exploration. In this paper, we propose a new histogram method for multidimensional data, called the Q-Histogram, based on the use of the quad-tree, which is a popular index structure for multidimensional data sets. The use of the compact representation of the target data obtainable from the quad-tree allows a fast construction of a histogram with the minimum number of scanning, i.e., only one scanning, of the underlying data. In addition to the advantage of computation time, the proposed method also provides a better performance than other existing methods with respect to the quality of selectivity estimation. We present a new measure of data skew for a histogram bucket, called the weighted bucket skew. Then, we provide an effective technique for skew-tolerant organization of histograms. Finally, we compare the accuracy and efficiency of the proposed method with other existing methods using both real-life data sets and synthetic data sets. The results of experiments show that the proposed method generally provides a better performance than other existing methods in terms of accuracy as well as computational efficiency.     

Information technology and productivity: Empirical evidence from the Chinese electronics industry

The importance of IT has been increasing in both developing and developed countries over the past three decades. However, almost all findings on IT productivity have been based on data collected in developed countries, while research on productivity in developing countries has been sparse. We studied the effect of IT investment on firm-level productivity in the Chinese electronics industry using a production function model. We found that the direction and size of the impact of IT investment on productivity in China were generally similar to results of studies for the United States and concluded that Chinese firms should invest more in IT. Finally, contrary to popular belief, labor is still an important factor in the production function of the Chinese electronics industry, despite its status as a high-technology industry.     

Combining models from neural networks and inductive learning algorithms

The knowledge-based artificial neural network (KBANN) is composed of phases involving the expression of domain knowledge, the abstraction of domain knowledge at neural networks, the training of neural networks, and finally, the extraction of rules from trained neural networks. The KBANN attempts to open up the neural network black box and generates symbolic rules with (approximately) the same predictive power as the neural network itself. An advantage of using KBANN is that the neural network considers the contribution of the inputs towards classification as a group, while rule-based algorithms like C5.0 measure the individual contribution of the inputs one at a time as the tree is grown. The knowledge consolidation model (KCM) combines the rules extracted using KBANN (NeuroRule), frequency matrix (which is similar to the Naïve Bayesian technique), and C5.0 algorithm. The KCM can effectively integrate multiple rule sets into one centralized knowledge base. The cumulative rules from other single models can improve overall performance as it can reduce error-term and increase R-square. The key idea in the KCM is to combine a number of classifiers such that the resulting combined system achieves higher classification accuracy and efficiency than the original single classifiers. The aim of KCM is to design a composite system that outperforms any individual classifier by pooling together the decisions of all classifiers. Another advantage of KCM is that it does not need the memory space to store the dataset as only extracted knowledge is necessary in build this integrated model. It can also reduce the costs from storage allocation, memory, and time schedule. In order to verify the feasibility and effectiveness of KCM, personal credit rating dataset provided by a local bank in Seoul, Republic of Korea is used in this study. The results from the tests show that the performance of KCM is superior to that of the other single models such as multiple discriminant analysis, logistic regression, frequency matrix, neural networks, decision trees, and NeuroRule. Moreover, our model is superior to a previous algorithm for the extraction of rules from general neural networks.     

Variants of the Uzawa method for saddle point problem

In this paper, we first propose three variants of the Uzawa method for solving the saddle point problem, and then we provide convergence results for the three proposed methods. Numerical experiments show that our proposed methods with three parameters perform about twice as fast as the GSOR (Generalized SOR) method with two parameters since the proposed methods have less workload per iteration than the GSOR.     

Effects of data set features on the performances of classification algorithms

As the need to analyze big data sets grows dramatically, the role that classification algorithms play in data mining techniques also increases. Big data analysis requires more of the data sets’ characteristics to be included, such as data structure, variety of sources, and the rate of update frequency. In this paper, we evaluate scenarios that examine which data set characteristics most affect the classification algorithms’ performance. It is still a complex issue to determine which algorithm is how strong or how weak in relation to which data set. Thus, our research experimentally examines how data set characteristics affect algorithm performance, both in terms of accuracy and in elapsed time. To do so, we use a multiple regression method to evaluate the causality between data set characteristics as independent variables, and performance metrics as dependent variables. We also examine the role that classification algorithms play as moderator in this causality. All benchmark data sets in a UCI database are used that are fit to run the classification algorithm. Based on the results of the experiment, we discuss the requirements of legacy classification algorithms to address big data analysis in a new business intelligence era.     

A sliding plate microgap rheometer for the simultaneous measurement of shear stress and first normal stress difference

A new generation of the "flexure-based microgap rheometer" (the N-FMR) has been developed which is also capable of measuring, in addition to the shear stress, the first normal stress difference of micrometer thin fluid films. This microgap rheometer with a translation system based on compound spring flexures measures the rheological properties of microliter samples of complex fluids confined in a plane couette configuration with gap distances of h = 1-400 μm up to shear rates of γ = 3000 s(-1). Feed back loop controlled precise positioning of the shearing surfaces with response times <1 ms enables to control the parallelism within 1.5 μrad and to maintain the gap distance within 20 nm. This precise gap control minimizes squeeze flow effects and allows therefore to measure the first normal stress difference N(1) of the thin film down to a micrometer gap distance, with a lower limit of N(1)/γ = 9.375×10(-11) η/h(2) that depends on the shear viscosity η and the squared inverse gap. Structural development of complex fluids in the confinement can be visualized by using a beam splitter on the shearing surface and a long working distance microscope. In summary, this new instrument allows to investigate the confinement dependent rheological and morphological evolution of micrometer thin films.     

Response selection and dynamic damper application to improve the identification of multiple input forces of narrow frequency band

In identifying input forces by using FRF matrices and response measurements, improper selection of the response measurement position and the excitation frequency, if it coincides with the natural frequencies of the structure, may cause doubtful accuracy and effectiveness. This paper presents two methods to improve force identification. The first is by improving the conditioning of the system FRF matrix by a proper selection of the measurement positions. An expression for the covariance matrix representing the uncertainty on the force estimate is defined. Then the response selection algorithm, which is both systematic and effective, is derived. The basic strategy in selecting the response positions is to let the smallest singular value be as large as possible. The second method is to modify the structure by attaching a dynamic damper at a suitable location to minimise the ill-conditional nature of the FRF matrix especially near a resonance frequency. The methods suggested are tested numerically and experimentally in two examples: a beam structure and a rectangular plate structure. The test results show that the response selection and the addition of a dynamic damper are very effective in solving ill-conditioning problems, especially when the structure has a large degree of freedom and when the excitation frequencies are near one of the structure's natural frequencies.     

Development of a radiative transfer model for the determination of toxic gases by Fourier transform–infrared spectroscopy with a support vector machine algorithm

Abstract

This report describes a radiative transfer model for Fourier transform-infrared (FT-IR) spectroscopy to create close-to-reality toxic gas spectra by reflecting the unique spectral responses of detectors and using the atmospheric radiative transfer code, MODTRAN. This system can be highly useful in overcoming the limitations for measuring toxic gases in open environments. The emulated gas spectra can be used to train support vector machine (SVM) for chemical gas detection. Its detection performance is evaluated with nerve agents (tabun, sarin, soman, and cyclosarin) and a simulant gas (sulfur hexafluoride) for indoor and outdoor experiments by using two off-the-shelf FT-IR gas detectors. The experimental results show that the proposed SVM algorithm successfully detected and classified targeted gases while reducing false negative and false positive detection rates.     

An experimental study on combustion efficiency of low grade anthracite in AFBC

This is the basic study to develop a fluidized bed combustion boiler which can use low grade anthracite. In this study, the anthracite of about 3400 kcal/kg was burned in the bench scale non-recycling atmospheric fluidized bed combustor of 200mm diameter and 2215mm height with the static bed height of 250mm and the combustion temperature range of 800–950°C. During the combustion, the effect of factors such as the superficial gas velocity in bed, the air ratio, the coal supply location and the coal particle size on the combustion efficiency, the elutriation ratio and the unburned carbon content both in elutriated ash and in drained ash was thoroughly analyzed. When the superficial gas velocity in bed is 0.7～2.2m/s, the air ratio is 1.0～1.6 and coal supply locations are 300,500 and 700mm above the air distributor, the combustion efficiencies range from 66% to 83.5% for the mean coal particle size of 0.209mm, and from 71% to 88% for the case of 0.265mm. The combustion efficiency decreases as the superficial gas velocity in bed and the air ratio increase. The lower the coal supply location is, the better the combustion efficiency becomes in general.     

A smart memory type of data acquisition system for shaft misalignment maintenance

This paper presents the development results of a smart memory module for data acquisition system to examine the characteristics of shaft misalignment. The smart memory module is a standalone type with high capacity, and it can communicate with computer via RS-232 interface in off-line after data acquisition. In the process of shaft misalignment maintenance, the misalignment data achieved by strain gauges on the turbine rotor are synchronized with the angle data gathered by 4 angular sensors boarded on the smart memory system. The real time data are acquisitioned in the manner of self operation during testing; then the data saved in smart memory are translated into a computer for data processing and analysis via RS-232C. For the better precision, the system uses two types of filters: a low pass filter on interface circuit and a transfer average filter in software. Also, the effectiveness of data acquisition operation using the developed smart memory module is proved through the experimental results using mock-up system.