Zeyu Zheng

Changes in Cross-Correlations as an Indicator for Systemic Risk

The 2008–2012 global financial crisis began with the global recession in December 2007 and exacerbated in September 2008, during which the U.S. stock markets lost 20% of value from its October 11 2007 peak. Various studies reported that financial crisis are associated with increase in both cross-correlations among stocks and stock indices and the level of systemic risk. In this paper, we study 10 different Dow Jones economic sector indexes, and applying principle component analysis (PCA) we demonstrate that the rate of increase in principle components with short 12-month time windows can be effectively used as an indicator of systemic risk—the larger the change of PC1, the higher the increase of systemic risk. Clearly, the higher the level of systemic risk, the more likely a financial crisis would occur in the near future.

Carbon-dioxide emissions trading and hierarchical structure in worldwide finance and commodities markets

In a highly interdependent economic world, the nature of relationships between financial entities is becoming an increasingly important area of study. Recently, many studies have shown the usefulness of minimal spanning trees (MST) in extracting interactions between financial entities. Here, we propose a modified MST network whose metric distance is defined in terms of cross-correlation coefficient absolute values, enabling the connections between anticorrelated entities to manifest properly. We investigate 69 daily time series, comprising three types of financial assets: 28 stock market indicators, 21 currency futures, and 20 commodity futures. We show that though the resulting MST network evolves over time, the financial assets of similar type tend to have connections which are stable over time. In addition, we find a characteristic time lag between the volatility time series of the stock market indicators and those of the EU CO(2) emission allowance (EUA) and crude oil futures (WTI). This time lag is given by the peak of the cross-correlation function of the volatility time series EUA (or WTI) with that of the stock market indicators, and is markedly different (>20 days) from 0, showing that the volatility of stock market indicators today can predict the volatility of EU emissions allowances and of crude oil in the near future.

Quantitative evaluation of the degree of sprout leaf bending of rice cultivars using P-type Fourier descriptors and principal component analysis

In quantitative studies of the shapes of plant organs, evaluation methods based on image analysis in terms of Fourier translation and principal component analysis have been recognized as important and useful methods. However, these methods cannot be used to evaluate the degree of bending of grass-type leaves. In this study, we provide a novel quantitative evaluation method that has been developed from a new type of Fourier translation named the P-type Fourier translation. It is an effective technique for evaluating the degree of bending of grass-type leaves. We have successfully applied this method to quantitatively evaluate the degree of bending of leaves of eight rice cultivars. The degree of bending of each rice leaf was evaluated in terms of several orthogonal principal component scores of P-type Fourier descriptors (F-PCs). It has been found in the present study that a few F-PCs are sufficient to quantitatively evaluate the essential variation among different rice cultivars. These inherent essential variations among rice cultivars are mainly described by only two individual F-PCs; the other variations such as the wilt caused by sunshine damage are described by another F-PC. The results also show that the F-PCs can somewhat differentiate the reasons for variation in the degree of bending. Because the F-PCs are independent from each other, they can be used as input data for QTL analysis.

Analysis of Realized Volatility in Two Trading Sessions of the Japanese Stock Market

We analyze realized volatilities constructed using high-frequency stock data on the Tokyo Stock Exchange. In order to avoid non-trading hours issue in volatility calculations we define two realized volatilities calculated separately in the two trading sessions of the Tokyo Stock Exchange, i.e. morning and afternoon sessions. After calculating the realized volatilities at various sampling frequencies we evaluate the bias from the microstructure noise as a function of sampling frequency. Taking into account of the bias to realized volatility we examine returns standardized by realized volatilities and confirm that price returns on the Tokyo Stock Exchange are described approximately by Gaussian time series with time-varying volatility, i.e. consistent with a mixture of distributions hypothesis.

Realized volatility and absolute return volatility: a comparison indicating market risk.

Measuring volatility in financial markets is a primary challenge in the theory and practice of risk management and is essential when developing investment strategies. Although the vast literature on the topic describes many different models, two nonparametric measurements have emerged and received wide use over the past decade: realized volatility and absolute return volatility. The former is strongly favored in the financial sector and the latter by econophysicists. We examine the memory and clustering features of these two methods and find that both enable strong predictions. We compare the two in detail and find that although realized volatility has a better short-term effect that allows predictions of near-future market behavior, absolute return volatility is easier to calculate and, as a risk indicator, has approximately the same sensitivity as realized volatility. Our detailed empirical analysis yields valuable guidelines for both researchers and market participants because it provides a significantly clearer comparison of the strengths and weaknesses of the two methods.

Early detection of liver disease using data visualisation and classification method

Abstract Detection of early-stage liver diseases is a challenge in medical field. Automated diagnostics based on machine learning therefore could be very important for liver tests of patients. This paper investigates 225 liver function test records (each record include 14 features), which is a subset from 1000 patients’ liver function test records that include the records of 25 patients with liver disease from a community hospital. We combine support vector data description (SVDD) with data visualisation techniques and the glowworm swarm optimisation (GSO) algorithm to improve diagnostic accuracy. The results show that the proposed method can achieve 96% sensitivity, 86.28% specificity, and 84.28% accuracy. The new method is thus well-suited for diagnosing early liver disease.

Establishment of a Quantitative Evaluation Method of Rice Plant Type Using P-type Fourier Descriptors

Abstract To evaluate rice plant type precisely at the seedling stage, we established a quantitative evaluation method using an image analysis. P-type Fourier descriptors, which could apply an “open-curve”, were used for the plant type, and the coefficients were summarized as the scores of principal components (PCs) by principal component analysis. At the same time, “conventional plant type traits”, leaf blade length, leaf blade angle and inter-leaf-blade length, were measured by traditional plant type measurement methods. Based on the PC scores, the plant type for each PC was reconstructed by inverse Fourier transformation and the morphological characteristics were evaluated. To examine whether our method could appropriately identify the characteristics of the varieties, we discriminated the varieties using the PC scores by support vector machines. The varieties were also discriminated using the conventional traits of plant type. The results indicated that both traits had equal discrimination efficiency. In addition, the combination of conventional traits and scores had the highest discrimination efficiency. The relationship between the PC scores and conventional traits was examined by multiple regression analysis. The PC scores were not correlated with size-related traits. From these results, our research clarified that a plant type evaluation method using P-type Fourier descriptors could evaluate rice plant type precisely combined with conventional methods.

An efficient clustering method for medical data applications

Clustering task is aimed at classifying elements into clusters, which is applied to different fields of the human activity. In this paper, an efficient clustering method by fast search and find of density peaks (FSFDP) is used for medical data applications. Different computing methods of the local density are compared and analyzed. For datasets composed by a small number of points, the local density might be affected by large statistical errors. Kernel local density is more accurate for estimating the density. Experiments were conducted to validate the efficiencies of the clustering method based on different local density for UCI benchmark and real-life datasets. The results show the feasibility and efficiency of the method for medical data clustering analysis.

Vibration analysis approach for corrosion pitting detection based on SVDD and PCA

This study is focused on corrosion pitting on the raceways and ball in rolling bearings. We analyze 224 records in the time domain, and combine support vector data description (SVDD) with principal component analysis (PCA) algorithm to improve diagnostic accuracy. Experiment results show that the proposed method can achieve good accuracy based on an imbalanced dataset. The new method is thus well-suited for corrosion pitting detection in rolling bearings.

Monitoring and fault diagnosis for industrial process

Monitoring and fault diagnosis are crucial for industrial process. In this paper, a simple and efficient manifold learning method is used for process monitoring and fault diagnosis. Firstly, local neighbor relationship of process data is used for process modelling, which divides process data into the embedding space and residual space. Then, different statistics and confidence limits are computed, which can be used for monitoring. Finally, the contribution analysis based on manifold learning is used for fault diagnosis. When the fault variables are found, quality control can be introduced to improve production safety and quality stabilization in industrial process. The manifold learning method is applied for one practical foods industrial production process. The experiment results show the feasibility and efficiency of the manifold learning method for monitoring and fault diagnosis.