区域经济预测模型的研究

本文介绍了几种基于定量分析的经济预测模型,提出了一个面向区域经济指标的组合预测模型。给出了确定组合模型中权系数的方法。研究结果表明组合预测是提高预测精度的有效方法,组合预测方法对区域经济的预测决策具有一定的现实意义。     

灰色神经网络模型应用于证券短期预测研究

证券市场中存在大量的非线性及混沌现象使得许多基于证券投资理论建议的各种线性模型在对其发展进行预测时往往得不到很好的结果。利用两种非线性预测模型并充分考虑各技术指标之间的序列相关性,在灰色GM(1,1)预测模型的基础上提出了组合灰色神经网络预测模型。通过对近期深证成指及首钢股份进行短期预测和检验,结果表明了该方法具有很高的精确度及广泛的应用前景。     

基于数据挖掘算法的变压器热点温度时序预测方法

为了解决油浸式电力变压器热点温度预测方法缺乏对短期热点温度走势的预测,无法满足动态增容决策要求的问题,以SFPSZ-180000/220型变压器为研究对象,首先,研究对比发现变压器的热点温度与负载率相关性最大,在此基础上,建立了基于支持向量回归的局部地区负荷预测模型,为变压器热点温度预测提供数据;其次,提出了基于数据挖掘算法的变压器热点温度时序预测方法,并在此基础上分别建立了支持向量回归、BP神经网络、决策树3种数据挖掘预测模型;最后,对一般输入-输出的建模方法的预测结果与基于时间延迟方法的预测结果,以及不同时间延迟下3种数据挖掘模型的预测结果进行了对比分析。结果表明,有外在输入的支持向量回归预测模型结果比BP神经网络和决策树吻合度更高,并且时间延时更小,预测结果精确度更高。有外在输入的支持向量回归预测模型在很大程度上提高了变压器热点温度预测精度,其预测结果可为变压器的动态增容决策提供有效参考。     

灰色马尔柯夫链在投资预测中的应用

本文针对电力工程项目投资受施工阶段性和施工干扰等因素影响,呈现随机波动性大的特点,将灰色GM(1,1)预测模型与马尔柯失链相结合,提出了灰色马尔柯夫链预测模型,并通过工程实例和与其它的传统预测方法相比验证,证明该模型预测精度高,适宜于随机波动性大的工程项目投资预测。     

特高含水水驱油田操作成本组合预测方法研究

操作成本作为油田生产过程中的主要组成部分,直接影响着油田的开发效益.为了不断降低操作成本以提高经济收益,以特高含水水驱油田为研究背景,通过该类油田操作成本的相关影响因素建立基于多重因素的两种预测模型——线性回归预测模型和非线性误差反传(BP)神经网络预测模型,最后利用组合预测法,将这两种预测模型组合在一起,以提高预测模型的有效性和可操作性,为特高含水水驱油田控制成本提供科学合理的依据.     

飞机易耗标准件维护需求预测的应用研究

飞机外场维护期间,易耗标准件种类和数量的维护很大程度依赖维护人员的经验,容易造成易耗标准件超量预备与缺失问题。对此,采用数理统计和回归分析的方法,以某型号飞机服役3年内的3种常用易耗标准件消耗数据为基础展开分析,求解得出飞机易耗标准件需求的消耗趋势和时间与消耗量的预测模型,并通过易耗标准件的实际消耗数据验证预测模型的符合性。该预测模型能根据飞机服役时间,有效预测易耗标准件特定种类的需求数量,降低工程维护和管理成本。     

上市公司财务困境动态预测模型的实证研究

本文利用数学方法对静态的财务指标进行动态调整,结合因子分析和Logistic回归分析,建立了上市公司财务困境的动态预测模型.结果表明,动态预测模型的预测精度远高于静态预测模型,因此可以预期其在财务困境预测领域应该具有广阔的应用空间.     

直线趋势预测中指数平滑法的计算机实现

使用直线趋势预测模型对现象的发展趋势进行预测，是统计预测中广泛使用的一种方法，然而在使用指数平没值确定直线趋势预测模型的待定参数时，始终两个难以解决的问题。文中对这两个问题进行了探讨，并提出了其计算机实现的程序。     

基于长短期记忆网络的轨道交通短期OD客流量预测

为了更准确地掌握轨道交通客流在线网中的时空分布,更高效地匹配客流需求与运输能力,实现提高轨道交通运输效率、改善运营服务质量的目的,提出了一种基于长短期记忆网络的短期OD(交通起止点)客流量预测方法。以历史客流数据为基础,定性分析车站间OD客流量的时空相关性,利用回归分析法定量分析客流影响因素,筛选出运营时刻、运营日特征、最低气温3个时间特征。为提高预测精度,以长短期记忆网络为基础,结合时间特征,为每对起讫点单独构建预测模型,形成了基于长短期记忆网络的轨道交通短期OD客流量预测模型。以苏州市为例进行验证,结果表明,加入了时间特征的短期OD客流量预测模型较移动平均模型、仅利用历史客流数据训练的基于LSTM网络的短期OD客流量预测模型,预测结果与真实值之间的误差降低了6.27%～8.58%,所提出的方法和模型可为轨道交通运营部门制定列车运行计划、组织客运工作提供更准确的数据资料。     

基于随机森林算法的陈垓灌区降水量预测模型

地区降雨量在短时间内的突然增大一般会导致不同程度的洪涝灾害,高效、精准地进行降雨量预测成为当下研究的热点问题。文章基于陈垓灌区梁山水文站实测降雨资料建立了基于随机森林的降水量预测模型,所建预测模型的各项指标均优于传统预测模型,可以高效、精准地预测该区域的降水量。     

The use for the competition theory of the industrial investment decisions—a case study of the Taiwan IC assembly industry

This study empirically analyzes model accuracy, and applies grey forecasting to handle non-linear problems, insufficient data resources and forecasting involving small samples, and to construct the co-opetition diffusion model for the Lotka–Volterra (L.V.) system. Furthermore, this study examines historical data comprising revenue trends in the Taiwanese IC assembly industry during the past ten years and selects from a range of forecasting models.Empirical study uses MAPE to precisely analyze revenue trends in the L.V. dynamic co-opetition diffusion model relation to the IC assembly industry. The nine companies will be selected from 4 to 11 of the modeling, the results of the LV model 64 accuracy test, its accuracy is higher than 95% accounted for 59 times, five times better than the grey prediction, showing LV competing diffusion model not only with grey prediction, and better than the traditional grey forecasting model to make a higher accuracy of the predicted value. Like grey forecasting, MAPE can promptly respond even given insufficient data. Additionally, MAPE is able to provide more accurate forecasting values than the traditional Grey forecasting model. This study demonstrates the applicability of the dynamic co-opetition theory forecasting model to the Taiwanese IC assembly industry and provides management with a reference for use in decisions aimed to increase managerial competitiveness.     

Fashion retail forecasting by evolutionary neural networks

Recent literature on nonlinear models has shown that neural networks are versatile tools for forecasting. However, the search for an ideal network structure is a complex task. Evolutionary computation is a promising global search approach for feature and model selection. In this paper, an evolutionary computation approach is proposed in searching for the ideal network structure for a forecasting system. Two years’ apparel sales data are used in the analysis. The optimized neural networks structure for the forecasting of apparel sales is developed. The performances of the models are compared with the basic fully connected neural networks and the traditional forecasting models. We find that the proposed algorithms are useful for fashion retail forecasting, and the performance of it is better than the traditional SARIMA model for products with features of low demand uncertainty and weak seasonal trends. It is applicable for fashion retailers to produce short-term retail forecasting for apparels, which share these features.     

A comparison of opinion and regression forecasting for an industrial product

Opinion-based forecasting techniques are widely used by industrial marketers. Rarely, however, are the results of these forecasts compared with alternative forecasting techniques and/or evaluated against actual operating results. In this study, opinion-based forecasting results for an industrial equipment manufacturer are evaluated against actual sales data. Further, the opinion-based predictions are compared with the predictions of a naive regression model. The results of these analyses suggest that the current opinion-based forecasting system is deficient.     

Forecasting New Product Adoption with Probabilistic Neural Networks

Forecasting the adoption of innovative products is an important managerial task. In this paper we examine the usefulness of a probabilistic neural network (PNN) algorithm for forecasting new product adoption. We compare this approach with one widely accepted forecasting procedure, the binomial logit model, and two other neural network algorithms: a feed‐forward neural network model estimated with backward propagation (NNBP), and a feed‐forward neural network model estimated with a genetic algorithm (NNGA). To test the relative forecasting accuracy of these algorithms, we examine the first‐time adoption of DVD players. Our analysis is based on longitudinal consumer data collected between March 2000 and March 2001. We find that the PNN algorithm significantly outperforms the logit model and the two remaining neural network algorithms.     

2种风电功率预测模型的比较

采用ARMA模型对风电功率进行了预测,并由ARMA方程推导出卡尔曼滤波状态方程和测量方程,从而将预测问题转化到状态空间,并利用卡尔曼滤波法预测了风电功率,比较了2种方法的预测效果。实例表明,卡尔曼滤波法能够提高风电功率的预测精度,并在一定程度上解决了时间序列分析法的预测时延问题,对电力系统的安全、稳定、经济运行以及提高运行效益具有重要意义。     

Forecasting the Market Penetration of New Technologies Using a Combination of Economic Cost and Diffusion Models

Forecasting market penetration is an essential step in the development, assessment, and commercialization of new technologies. Among the many forecasting approaches available are the economic cost model and the diffusion model. Separately, each of these approaches has been used in many applications of market penetration forecasting. In this article, A. P. S. Teotia and P. S. Raju briefly review these two approaches and then describe a methodology for forecasting market penetration using both of these approaches sequentially. They illustrate this approach with an example of market penetration forecasting for energy-efficient electric motors. A combination approach, which incorporates the strengths of two or more approaches, may be superior in many instances to the use of any one approach alone.     

灰色模型在股票价格预测中的应用——以中国石化为例

通过对原始数据进行筛选,将GM(1,1)模型用于股票预测,对交易日收盘价格进行预测,并以中国石化(600028)2012-01-04至2012-12-31的交易数据为例进行了分析。分析结果表明,灰色预测模型的平均预测准确度为98.63%。考虑到股票交易规则,42%的预测数据为有效预测,有效预测的平均预测准确度为99.31%。     

Kurzfrist-Lastprognose für die Day-Ahead- und Intraday-Stromnachfrage bei sich systematisch ändernden Lasten

Load forecasts are used in various fields of the German energy economic to plan and to optimize the schedule of the power generation or the purchase of power from the markets based on the results of the forecasts. Therefor accurate load forecasts are necessary. But many load forecasting models reach their limits when dealing with systematic changes in the profile of the energy demand, since the model is usually calibrated by historic data so the relation between the load and the input parameters are estimated. Due to changes in the load profile the load level is moving to another level compared to the historic one. While the forecasting model is still calibrated on the old level, this can lead to higher forecasting errors and these can in turn have negative consequences on the following optimization steps. That is why a methodological approach is presented so that the forecasting model is able to adapt a systematic change in the load profile. Therefor the presented approach is at first applied to a case of application, before it is applied to two more extreme variations of the load profile to identify possible limits of the presented approach.     

Developing and Using Forecasting Models of Consumer Durables The Case of Color Television

Predicting the future sales of new and established products is a critical activity for companies to be able to plan and control their operations. Forecasting consumer durable sales is an especially difficult and challenging task since the marketplace is always changing. Barry Bayus, Saman Hong and Russell Labe present the concept of a market-driven forecasting model and discuss an application involving the forecasting of color television industry sales for RCA's Consumer Electronics Division. An important aspect of this application is that a single approach or model was inadequate to accurately forecast sales over the entire period since the introduction of color TV. Econometric and simulation models which were developed are described, along with their forecasting performance and management acceptance and use.     

Robust Testing of the Utility‐based High‐technology Product Sales Forecasting Models Proposed by Decker and Gnibba‐Yukawa (2010)

Decker and Gnibba‐Yukawa (2010) propose an elegant utility‐based model for forecasting the sales of high‐technology products and suggest that the model yields forecasts that are highly accurate. However, this finding is based on forecasts for a total of only six holdout observations shared across three products. This number of observations is insufficient for reliable inferences to be drawn about the accuracy of a method and the use of such a small data set runs counter to an accepted principle of forecast evaluation. The authors’ proposed model was tested on more extensive data and sensitivity analysis applied to the results. No evidence was found that the utility‐based model could outperform a relatively simple extrapolative model despite the much greater effort involved in applying the proposed model. In addition, the utility‐based model is only applicable for forecasting sales during a narrow interval in a product's life cycle and requires several periods of historic sales data before it can be implemented. It also depends heavily on the accurate estimates of parameters that are determined outside the model (and which may depend on difficult judgments by managers) and assumes that consumers or households will only purchase the product once between the launch date and the forecast horizon. In light of this, it is argued that the utility‐based model is likely to have limited usefulness as a sales forecasting tool.