Comparing classification accuracy of neural networks,binary logit regression and discriminant analysis for insolvency prediction of life insurers

Past studies have documented the failure of the Insurance Regulatory Information System (IRIS) to provide adequate warning of insurer financial distress or insolvency. As a result, scholars have examined alternative parametric and non-parametric models to predict insurer insolvency. This study uses a neural network, a non-parametric alternative to past techniques, and shows how this methodology predicts insurer insolvency more effectively than parametric models.     

基于神经网络方法的风机故障诊断

针对旋转机械故障征兆与故障模式映射的复杂性,将反向传播(BP)网络、径向基(RBF)网络和概率网络(PNN)用于风机进行故障诊断,并比较了3种网络的诊断精度。以风机振动信号的7段频谱能量峰值作为故障特征,采用训练好的神经网络进行故障辨识,结果表明,RBF网识别精度高于PNN网络,BP网络表现较差。     

A combined machine learning algorithms and DEA method for measuring and predicting the efficiency of Chinese manufacturing listed companies

Data Envelopment Analysis (DEA) is a linear programming methodology for measuring the efficiency of Decision Making Units (DMUs) to improve organizational performance in the private and public sectors. However, if a new DMU needs to be known its efficiency score, the DEA analysis would have to be re-conducted, especially nowadays, datasets from many fields have been growing rapidly in the real world, which will need a huge amount of computation. Following the previous studies, this paper aims to establish a linkage between the DEA method and machine learning (ML) algorithms, and proposes an alternative way that combines DEA with ML (ML-DEA) algorithms to measure and predict the DEA efficiency of DMUs. Four ML-DEA algorithms are discussed, namely DEA-CCR model combined with back-propagation neural network (BPNN-DEA), with genetic algorithm (GA) integrated with back-propagation neural network (GANN-DEA), with support vector machines (SVM-DEA), and with improved support vector machines (ISVM-DEA), respectively. To illustrate the applicability of above models, the performance of Chinese manufacturing listed companies in 2016 is measured, predicted and compared with the DEA efficiency scores obtained by the DEA-CCR model. The empirical results show that the average accuracy of the predicted efficiency of DMUs is about 94%, and the comprehensive performance order of four ML-DEA algorithms ranked from good to poor is GANN-DEA, BPNN-DEA, ISVM-DEA, and SVM-DEA.     

Investigating the accuracy of cross-learning time series forecasting methods

The M4 competition identified innovative forecasting methods, advancing the theory and practice of forecasting. One of the most promising innovations of M4 was the utilization of cross-learning approaches that allow models to learn from multiple series how to accurately predict individual ones. In this paper, we investigate the potential of cross-learning by developing various neural network models that adopt such an approach, and we compare their accuracy to that of traditional models that are trained in a series-by-series fashion. Our empirical evaluation, which is based on the M4 monthly data, confirms that cross-learning is a promising alternative to traditional forecasting, at least when appropriate strategies for extracting information from large, diverse time series data sets are considered. Ways of combining traditional with cross-learning methods are also examined in order to initiate further research in the field.     

Financial distress prediction based on SVM and MDA methods: the case of Chinese listed companies

How to accurately predict financial distress is an important issue for enterprise managers, investors, creditors and supervisors. In this paper we develop SVM models (Support Vector Machine) and MDA (Multivariate Discriminant Analysis) models, using Chinese listed companies as our sample. The empirical results show that the prediction ability of SVM models outperforms the MDA models. Additionally, internal governance and external market variables, as well as macroeconomic variables are added as the predictive variables. The results indicate that these variables have theoretical and empirical linkage with the financial distress of Chinese listed companies.     

基于PCA的CS-RBF网络的股票预测研究

针对传统的RBF神经网络在选取中心矢量参数时的不足,提出用具有较强跳出局部最优的布谷鸟算法(CS)优化RBF神经网络的中心矢量的改进算法,并将该算法应用于股票价格的预测,仿真结果表明:该算法的预测精度比传统的RBF算法的预测精度高,是一种有效的股票预测方法。     

基于混沌理论和PSO神经网络的短时交通流预测

交通流预测已成为智能交通的重要组成部分,针对短时交通流的非线性和不确定性,文中根据实际交通流中存在的混沌,利用C-C方法和小数据量法对交通流混沌进行了分析,在交通流混沌时间序列相空间重构的基础上构建了基于粒子群优化神经网络的单点单步预测模型,运用该模型对实际采集的美国加州城市快速路交通流数据进行了仿真研究,结果表明,该预测模型具有较高的预测精度,能够满足智能交通控制和诱导的需求。     

Does audit report information improve financial distress prediction over Altman's traditional Z‐Score model?

We analyze empirically the usefulness of combining accounting and auditing data in order to predict corporate financial distress. Concretely, we examine whether audit report information incrementally predicts distress over a traditional accounting model: the Altman's Z‐Score model. Although the audit report seems to play a critical part in financial distress prediction because auditors should warn investors about any default risks, this is the first study that uses audit report disclosures for predicting purposes. From a dataset of 1,821 Spanish distressed private firms, we analyze a sample of distressed and non‐distressed firms and develop logit prediction models. Our results show that while the only accounting model registers a classification accuracy of 77%, combined models of accounting and auditing data exhibit considerably higher accuracy (about 87%). Specifically, our findings indicate that the number of disclosures included in the audit report, as well as disclosures related to a firm's going concern status, firms’ assets, and firms’ recognition of revenues and expenses contribute the most to the prediction. Our empirical evidence has implications for financial distress practice. For managers, our study highlights the importance of audit report disclosures for anticipating a financial distress situation. For regulators and auditors, our study underscores the importance of recent changes in regulation worldwide intended to increase auditor's transparency through a more informative audit report.     

Systemic financial risk early warning of financial market in China using Attention-LSTM model

We propose an Attention-LSTM neural network model to study the systemic risk early warning of China. Based on text mining, the network public opinion index is constructed and used as a training set to be incorporated into the early warning model to test the early warning effect. The results show that: (i) the network public opinion is the non-linear Granger causality of systemic risk. (ii) The Attention-LSTM neural network has strong generalization ability. Early warning effects have been significantly improved. (iii) Compared with the BP neural network model, the SVR model and the ARIMA model, the LSTM neural network early warning model has a higher accuracy rate, and its average prediction accuracy for systemic risk indicators has been improved over short, medium and long terms. When the attention mechanism is included in the LSTM, the Attention-LSTM neural network model is even more accurate in all the cases.     

The tourism forecasting competition

We evaluate the performances of various methods for forecasting tourism data. The data used include 366 monthly series, 427 quarterly series and 518 annual series, all supplied to us by either tourism bodies or academics who had used them in previous tourism forecasting studies. The forecasting methods implemented in the competition are univariate and multivariate time series approaches, and econometric models. This forecasting competition differs from previous competitions in several ways: (i) we concentrate on tourism data only; (ii) we include approaches with explanatory variables; (iii) we evaluate the forecast interval coverage as well as the point forecast accuracy; (iv) we observe the effect of temporal aggregation on the forecasting accuracy; and (v) we consider the mean absolute scaled error as an alternative forecasting accuracy measure. We find that pure time series approaches provide more accurate forecasts for tourism data than models with explanatory variables. For seasonal data we implement three fully automated pure time series algorithms that generate accurate point forecasts, and two of these also produce forecast coverage probabilities which are satisfactorily close to the nominal rates. For annual data we find that Naïve forecasts are hard to beat.     

Structural analysis and forecast of gold price returns

Gold has multiple attributes and its price is affected by various factors in the market. This paper studies the dynamic relationship between the gold price returns and its affecting factors. Then we use the STL-ETS, neural network and Bayesian structural time series model to predict the gold price returns, and compare their performance with the benchmark models. The results show that the shocks of crude oil returns and VIX have the positive effect on gold price returns, the shocks of the US dollar index have the negative effect on gold price returns. And the fluctuation of gold price returns mainly depends on crude oil price returns shocks. STL-ETS model can accurately fit the fluctuation trend of the gold price returns and improve prediction accuracy.     

基于RBF神经网络的船舶辐射噪声预测

本文根据船舶辐射噪声信号的混沌特性,基于重构相空间的轨迹演化规律,构建一种快速的RBF神经网络对船舶辐射噪声进行预测。实验证明,运用这种RBF神经网络对船舶辐射噪声的预测精度高于Volterra自适应滤波预测滤波器,而且更快的收敛速度。     

Multivariate Small Area Estimation of Multidimensional Latent Economic Well-being Indicators

Factor analysis models are used in data dimensionality reduction problems where the variability among observed variables can be described through a smaller number of unobserved latent variables. This approach is often used to estimate the multidimensionality of well-being. We employ factor analysis models and use multivariate empirical best linear unbiased predictor (EBLUP) under a unit-level small area estimation approach to predict a vector of means of factor scores representing well-being for small areas. We compare this approach with the standard approach whereby we use small area estimation (univariate and multivariate) to estimate a dashboard of EBLUPs of the means of the original variables and then averaged. Our simulation study shows that the use of factor scores provides estimates with lower variability than weighted and simple averages of standardised multivariate EBLUPs and univariate EBLUPs. Moreover, we find that when the correlation in the observed data is taken into account before small area estimates are computed, multivariate modelling does not provide large improvements in the precision of the estimates over the univariate modelling. We close with an application using the European Union Statistics on Income and Living Conditions data.     

企业财务预警的研究方法及其改进——基于文献综述

企业财务预警研究方法经历了趋势分析、判别分析、人工智能技术、传统方法的改进和前沿技术的采用四个发展阶段.作为方法的改进,基于聚类、粗糙集、神经网络的财务困境预警方法更具有科学性.     

Nonlinear Correlograms and Partial Autocorrelograms*

This paper proposes neural network‐based measures of predictability in conditional mean, and then uses them to construct nonlinear analogues to autocorrelograms and partial autocorrelograms. In contrast to other measures of nonlinear dependence that rely on nonparametric estimation of densities or multivariate integration, our autocorrelograms are simple to calculate and appear to work well in relatively small samples.     

基于深度残差网络的刀具磨损量预测方法研究

数控机床刀具磨损将直接影响着加工产品的精度,导致产品质量下降。在自动化程度越来越高的数控加工中,监测刀具状态变得更加困难。为了保证产品质量,快速、精确的预测刀具磨损量,本文提出基于深度残差神经网络的多传感器刀具磨损量预测方法,首先,该方法提取振动、切削力和声发射传感器信号的时域特征,然后,利用深度残差网络对时域特征进行监督学习训练,最后,把训练好的模型对测试数据进行测试。本文提出的模型通过试验验证其有效性,模型评价指标绝对均值(MAE)约为1.51×10-3mm,具有较高的预测精度。     

基于人工神经网络的互联网公司估值方法研究

资本市场认为互联网公司市值的驱动因素应包括盈利因子、运营因子、流量因子和协同因子。将协同效应指标考虑到公司估值体系中,意图构造互联网公司优化估值模型。使用美股上市的互联网企业数据建立了评价指标体系,通过因子分析实现了二级指标降维,通过实证分析确认了四个因子与公司市值的相关关系,最后构建了基于人工神经网络BP算法的互联网公司估值模型,通过预测数据的检验发现模型的准确度较高。随着2018年互联网公司美股上市潮的持续,该模型能有效为资本市场估值提供参考。     

多属性多参数横波预测

岩石物理理论模型是地震岩石物理研究的基础,通过对叠前反演中横波速度的应用,提出了横波速度曲线优化的估算方法,通过对PNN视神经网络算法以及非线性多属性参数的分析,解决了横波预测中误差较大的影响,研究结果为后续的叠前反演打下了很好的基础。     

Machine learning model for Bitcoin exchange rate prediction using economic and technology determinants

In recent years, Bitcoin exchange rate prediction has attracted the interest of researchers and investors. Some studies have used traditional statistical and econometric methods to understand the economic and technology determinants of Bitcoin, few have considered the development of predictive models using these determinants. In this study, we developed a two-stage approach for exploring whether the information hidden in economic and technology determinants can accurately predict the Bitcoin exchange rate. In the first stage, two nonlinear feature selection methods comprising an artificial neural network and random forest are used to reduce the subset of potential predictors by measuring the importance of economic and technology factors. In the second stage, the potential predictors are integrated into long short-term memory (LSTM) to predict the Bitcoin exchange rate regardless of the previous exchange rate. Our results showed that by using the economic and technology determinants, LSTM could achieve better predictive performance than the autoregressive integrated moving average, support vector regression, adaptive network fuzzy inference system, and LSTM methods, which all use the previous exchange rate. Thus, information obtained from economic and technology determinants is more important for predicting the Bitcoin exchange rate than the previous exchange rate.     

Analysis of the gold fixing price fluctuation in different times based on the directed weighted networks

According to the complex network theory, this paper constructs the gold fixing price fluctuation directed weighted network (GFPFDWN) at 10:30 a.m. (A.M.) and 15:00p.m. (P.M.) London Greenwich Mean Time, and studies the law of the gold fixing price fluctuation by analyzing the basic statistical data and characteristics of the GFPFDWN. The results show that the abnormal distribution of the gold fixing price (GFP) at A.M. and P.M. is confirmed by the statistics, the core fluctuation state of the GFPFDWN is reflected in the first 1.8% nodes, and most of the nodes have smaller strength, only a few nodes have larger strength, which has the characteristics of a typical scale-free network. Meanwhile, the nodes with a large strength are closely related among them, which must appear earlier, but the nodes appearing early are not necessarily the nodes with a large strength. The nodes of the GFPFDWN present a short-range correlation in different periods, and the cumulative time of the new nodes shows a high linear growth trend. In addition, the nodes of the GFPFDWN show the characteristics with a small betweenness, clustering coefficient and node strength in different periods, which are different from the characteristics of the random network and chaotic network. When these nodes with small strength appear, which means that this period is in a transitional period, then identifying and analyzing these nodes can effectively predict the fluctuation of the gold fixing price in the next period.