基于小波自适应支持向量机的短期电价预测

电价预测对于发电商、供电企业以及市场监管者都具有重要的意义。提出一种小波自适应支持向量机预测模型,先将电价时间序列作小波分解得到低频和高频分量,再采用自适应调整法,自动地为支持向量机选择较好的参数对电价小波分量逐一预测,最后通过小波重构得到电价最终预测结果。实例证明前述方法得到的预测精度高于BP、RBF、SVM等传统预测模型。     

中国粮食供需的预测指标体系及模型设计

通过建立粮食需求预测指标体系,从口粮、饲料粮、种子粮、工业用粮及粮食损耗角度实现了粮食需求预测。并采用基于三次指数平滑模型、灰色预测模型、支持向量机预测模型的组合预测模型,成功实现了粮食供给预测。最后,在粮食供需综合分析中,确认了粮食供需缺口的存在性。     

基于KPCA-SVM的公路客运量预测研究

支持向量机(Support Vector Machine,SVM)是建立在统计学习理论和结构风险最小化准则基础上的机器学习方法,该方法可以较好的解决以往很多学习方法的小样本、高维数、非线性和局部最小点等实际问题.本文利用支持向量机(SVM)回归理论和方法,建立基于核函数主成分支持向量机(Kernel Principal Component Analysis-Support Vector Machine,KPCA-SVM)回归模型,并用2000-2008年杭州市公路客运量为样本进行了预测,结果表明,KPCA-SVM模型具有较高的预测精度和可靠性,是一种有效的公路客运量预测方法.     

Do money and financial variables help forecasting output in emerging European Economies?

Whether including monetary aggregates and different financial variables into small scale BVAR models improves the accuracy of output forecasts is tested for three emerging European economies. Various specifications for the priors of the BVAR models are used. The results are found to vary with respect to prior specification, variables, as well as prediction horizon. The evidence is stronger when the forecasting accuracy is compared based on log predictive likelihood but weaker when the RMSEs are used. These results may constitute evidence against dismissing the monetary aggregates or financial variables as completely irrelevant.     

大数据背景下基于ABC-SVM的建筑工程造价预测

建筑工程项目决策阶段信息量少,精准高效的造价预测是科学决策的关键.为了提高项目前期工程造价预测的精度,探讨如何利用历史项目大数据及机器学习进行新建建筑工程项目的造价预测至关重要.本文首先通过文献研究确定了建筑工程决策阶段造价的主要影响因素,然后利用人工蜂群算法(ABC)对支持向量机(SVM)参数即惩罚因子C和核函数参数g进行优化计算,最终构建了基于ABC-SVM的建筑工程造价预测模型.最后以某工程造价数据平台上的84个建筑工程项目为数据源进行模型验证,结果显示,与GRID-SVM模型和BP神经网络模型相比,本文所提的ABC-SVM模型的预测精度更高,具有更好的适用性.     

Dynamics of inductive inference in a unified framework

We present a model of inductive inference that includes, as special cases, Bayesian reasoning, case-based reasoning, and rule-based reasoning. This unified framework allows us to examine how the various modes of inductive inference can be combined and how their relative weights change endogenously. For example, we establish conditions under which an agent who does not know the structure of the data generating process will decrease, over the course of her reasoning, the weight of credence put on Bayesian vs. non-Bayesian reasoning. We illustrate circumstances under which probabilistic models are used until an unexpected outcome occurs, whereupon the agent resorts to more basic reasoning techniques, such as case-based and rule-based reasoning, until enough data are gathered to formulate a new probabilistic model.     

Uncertainty and crude oil market volatility: new evidence

ABSTRACT

The main goal of this paper is to investigate the predictability of five economic uncertainty indices for oil price volatility in a changing world. We employ the standard predictive regression framework, several model combination approaches, as well as two prevailing model shrinkage methods to evaluate the performances of the uncertainty indices. The empirical results based on simple autoregression models including only one index suggest that global economic policy uncertainty (GEPU) and US equity market volatility (EMV) indices have significant predictive power for crude oil market volatility. In addition, the model combination approaches adopted in this paper can improve slightly the performances of individual autoregressive models. Lastly, the two model shrinkage methods, namely Elastin net and Lasso, outperform other individual AR-type model and combination models in most forecasting cases. Other empirical results based on alternative forecasting methods, estimation window sizes, high/low volatility and economic expansion/recession time periods further make sure the robustness of our major conclusions. The findings in this paper also have several important economic implications for oil investors.     

A novel time-series model based on empirical mode decomposition for forecasting TAIEX

Stock price prediction is regarded as a challenging task of the financial time series prediction process. Time series models have successfully solved prediction problems in many domains, including the stock market. Unfortunately, there are two major drawbacks in stock market by time-series model: (1) some models cannot be applied to the datasets that do not follow the statistical assumptions; and (2) most time-series models which use stock data with many noises involutedly (caused by changes in market conditions and environments) would reduce the forecasting performance. For solving the above problems and promoting the forecasting performance of time-series models, this paper proposes a hybrid time-series support vector regression (SVR) model based on empirical mode decomposition (EMD) to forecast stock price for Taiwan stock exchange capitalization weighted stock index (TAIEX). In order to evaluate the forecasting performances, the proposed model is compared with autoregressive (AR) model and SVR model. The experimental results show that the proposed model is superior to the listing models in terms of root mean squared error (RMSE). And the more fluctuation year (2000–2001) occurs, the better accuracy of proposed model will be obtained.     

Bankruptcy prediction for small- and medium-sized companies using severely imbalanced datasets

Bankruptcy prediction is still important topic receiving notable attention. Information about an imminent bankruptcy threat is a crucial aspect of the decision-making process of managers, financial institutions, and government agencies. In this paper, we utilize a newly acquired dataset comprising financial parameters derived from the annual reports of small- and medium-sized companies. The data, which reveal the true ratio between bankrupt and non-bankrupt companies, are severely imbalanced and only contain a small fraction of bankrupt companies. Our solution to overcome this challenging scenario of imbalanced learning was to adopt three one-class classification methods: a least-squares approach to anomaly detection, an isolation forest, and one-class support vector machines for comparison with conventional support vector machines. We provide a comprehensive analysis of the financial attributes and identify those that are most relevant to bankruptcy prediction. The highest prediction performance in terms of the geometric mean score is 91%. The results are validated on two datasets from the manufacturing and construction industries.     

Predicting stock realized variance based on an asymmetric robust regression approach

This paper introduces an asymmetric robust weighted least squares (ARLS) approach to improve the forecasting performance of the heterogeneous autoregressive model for realized volatility. The ARLS approach down-weights extreme observations to limit the bad influence of outliers on the estimated parameters. Compared with existing robust regression methods, our model further takes into account the asymmetry of outliers using a class of kernel functions. Out-of-sample results show the ARLS approach can generate more accurate forecasts of the S&P 500 index realized volatility in the statistical and economic senses. The model that considers the asymmetry of outliers gains superior performance among various robust regression competitors. The forecasting improvements also hold in other international stock markets. More importantly, the source of the predictive ability of the ARLS model comes from the less biased and more efficient parameter estimation.     

Does the vector error correction model perform better than others in forecasting stock price? An application of residual income valuation theory

This paper employs a multi-equation model approach to consider three statistic problems (heteroskedasticity, endogeneity and persistency), which are sources of bias and inefficiency in the predictive regression models. This paper applied the residual income valuation model (RIM) proposed by Ohlson (1995) to forecast stock prices for Taiwan three sectors. We compare relative forecasting accuracy of vector error correction model (VECM) with the vector autoregressive model (VAR) as well as OLS and RW models used in the prior studies. We conduct out-of-sample forecasting and employ two instruments to assess forecasting performance. Our empirical results suggest that the VECM statistically outperforms other three models in forecasting stock prices. When forecasting horizons extend longer, VECM produces smaller forecasting errors and performs substantially better than VAR, suggesting that the ability of VECM to improve VAR forecast accuracy is stronger with longer horizons. These findings imply that an error correction term (ECT) of the VECM contributes to improving forecast accuracy of stock prices. Our economic significance analyses and robustness tests for different data frequency are in favor of the superiority of VECM estimator.     

Ideology and prediction

To the formative giants of social science Hegel, Marx, Pareto, Weber, Freud, and Lewin the prediction task was central for both theory-building and practical benefits to society. Basic to their thinking was a preoccupation with the conflict and interaction of ideological forces and alignments behind dialectical processes determining social change and the course of much of the human future. Today their thinking largely lies fallow within both social science and forecasting research and practice. This is a report of a modern attempt to investigate, reactivate, and advance their insights in modern empirical terms. Results to date include a recent study of political, economic, social, and both short-term and long-term forecasting that indicates considerable predictive powers for modern forecasting approaches based on the measurement of ideological variables in surveys and the use of a matrix format for the analysis of social change and the projection of futures.     

Copper Price Behaviour and the London Metal Exchange

This paper examines how variables which describe the expectations of consumers can contribute to the explanation of observed expenditure patterns and how measured series of such expectations can be used in a forecasting model to improve the prediction of short-term consumer expenditures. The expectations data are based on the British Market Research Bureau's Financial Expectations Survey and the respective series that are derived are tested in correlation and regression exercises against quarterly aggregate consumer expenditure series. The exercise finds that the information contained in these financial expectations has significant value for predicting expenditures in the period 1 to 12 months ahead. The forecasting models based on the expectational data generally perform as well as those based on conventional economic variables and the leading indicator properties of the expectations, combined with their rapid availability, enhance their value as a potential source of forecasting information.     

Validity of analogical predictions

Comparison-Based Prediction (CBP) is a formalized method for using analogical reasoning to generate estimates early in conceptual design and planning. Two validation studies were performed, and both showed very high predictive validity when performance data were available for the comparison systems and much lower predictive validity when the comparison system data had to be estimated. The CBP method has also been found to reduce standard deviations by 25% or more. The method appears to have a wide range of applicability for prediction tasks.     

Out‐of‐Sample Exchange Rate Forecasting and Macroeconomic Fundamentals: The Case of Japan

This study explores the respective out‐of‐sample exchange rate forecasting abilities of five macroeconomic fundamental models in comparison to a naïve random walk model for Japan during the post‐Bretton Woods era. To assess the influence of major economic changes, we estimate both linear and nonlinear models for all the macroeconomic fundamentals. Overall, most structural exchange rate models outperform a naïve random walk model in terms of forecasting accuracy in the short horizon. When the fundamentals are only linearly modelled, the forecasting ability of the Taylor rule is generally superior to other fundamental models. When the fundamentals are nonlinearly specified, the predictability of some other models rises dramatically to match that of the Taylor rule models in short and/or long horizons. Of importance, we determine that the yen/dollar exchange rate forecasting performance effectively improves in several fundamental models when influential economic changes are incorporated.     

A Litterman BVAR approach for production forecasting of technology industries

Forecasting the production of technology industries is important to entrepreneurs and governments, but usually suffers from market fluctuation and explosion. This paper aims to propose a Litterman Bayesian vector autoregression (LBVAR) model for production prediction based on the interaction of industrial clusters. Related industries within industrial clusters are included into the LBVAR model to provide more accurate predictions. The LBVAR model possesses the superiority of Bayesian statistics in small sample forecasting and holds the dynamic property of the vector autoregression (VAR) model. Two technology industries in Taiwan, the photonics industry and semiconductor industry are used to examine the LBVAR model using a rolling forecasting procedure. As a result, the LBVAR model was found to be capable of providing outstanding predictions for these two technology industries in comparison to the autoregression (AR) model and VAR model.     

基于主成分分析和支持向量机的个人信用评估

本文针对信用评估指标维数较高的问题,运用主成分分析与支持向量机理论建立了一个新的个人信用评估预测模型。为反映该模型在信用评估分类方面的优越性,又分别建立了基于神经网络、K近邻判别分析等多种理论的信用评估模型,并用同一组数据对不同的模型分别进行训练,然后比较其预测分类正确率。实验结果表明,基于主成分分析与支持向量机理论的个人信用评估模型具有较优的预测分类正确率。     

Forecast accuracy of small and large scale dynamic factor models in developing economies

Developing economies usually present limitations in the availability of economic data. This constraint may affect the capacity of dynamic factor models to summarize large amounts of information into latent factors that reflect macroeconomic performance. This paper addresses this issue by comparing the accuracy of two kinds of dynamic factor models at GDP forecasting for six Latin American countries. Each model is based on a dataset of different dimensions: a large dataset composed of series belonging to several macroeconomic categories (large scale dynamic factor model) and a small dataset with a few prescreened variables considered as the most representative ones (small scale dynamic factor model). Short‐term pseudo real time out‐of‐sample forecast of GDP growth is carried out with both models reproducing the real time situation of data accessibility derived from the publication lags of the series in each country. Results (i) confirm the important role of the inclusion of latest released data in the forecast accuracy of both models, (ii) show better precision of predictions based on factors with respect to autoregressive models and (iii) identify the most adequate model for each country according to availability of the observed data.     

Combination forecasts of tourism demand with machine learning models

The main objective of this study is to analyse whether the combination of regional predictions generated with machine learning (ML) models leads to improved forecast accuracy. With this aim, we construct one set of forecasts by estimating models on the aggregate series, another set by using the same models to forecast the individual series prior to aggregation, and then we compare the accuracy of both approaches. We use three ML techniques: support vector regression, Gaussian process regression and neural network models. We use an autoregressive moving average model as a benchmark. We find that ML methods improve their forecasting performance with respect to the benchmark as forecast horizons increase, suggesting the suitability of these techniques for mid- and long-term forecasting. In spite of the fact that the disaggregated approach yields more accurate predictions, the improvement over the benchmark occurs for shorter forecast horizons with the direct approach.     

Bayesian forecasting of federal funds target rate decisions

In this paper we examine which macroeconomic and financial variables have most predictive ability for the federal funds target rate decisions made by the Federal Open Market Committee (FOMC). We conduct the analysis for the 157 FOMC decisions during the period January 1990–June 2008, using dynamic ordered probit models with a Bayesian endogenous variable selection methodology and real-time data for a set of 33 candidate predictor variables. We find that indicators of economic activity and forward-looking term structure variables, as well as survey measures are most informative from a forecasting perspective. For the full sample period, in-sample probability forecasts achieve a hit rate of 90%. Based on out-of-sample forecasts for the period January 2001–June 2008, 82% of the FOMC decisions are predicted correctly.