基于企业数据的预测与评价方法研究

发布时间：2018-01-02 06:44

本文关键词：基于企业数据的预测与评价方法研究　出处：《合肥工业大学》2016年博士论文　论文类型：学位论文

【摘要】：随着科学技术的发展,尤其是互联网和计算机通信等技术的进步与深入应用,人们已经不可避免地身处在数据海洋里,无论是日常生活还是企业的生产经营活动,数据都无时无刻地产生着。数据的产生渠道不断地增加,使得人们获取数据的方式变得越来越容易,相应的成本也不断地降低,但是同样地,数据量的增加和数据形态的日益复杂导致了数据的利用变得越来越困难,如何有效处理和利用数据,使数据发挥出其应有的作用,为人类生活和社会发展提供强大的支持等问题已成为当前社会各界讨论的焦点。随着数据问题的不断升温,全球学术界、工业界甚至各国政府都高度关注相关的数据技术,数据科学也逐渐成为了一门新的研究科学。由于数据的合理利用不仅可以服务大众,而且可以为企业带来巨大的商业价值,因此数据被称为“未来的新石油”,未来的竞争焦点将聚焦于对数据的占有、控制和运用上。虽然数据的重要性得到了世界的公认,但是由于社会发展的不平衡,国内许多企业仍然无法享受数据为其带来的利益,这些企业在数据的使用上仍然处于初始阶段,相应的企业数据仍然保持着“沉睡”的状态。本文在广泛收集和综合分析国内外企业对数据处理和使用过程中各方面研究成果的基础上,对相应企业进行了深入调研,结合企业目前在数据处理和使用过程中存在的一些问题,确定了本文的研究方向。首先,针对企业数据复杂度和数据维度不断增加、企业数据应用困难等问题,本文对相关的数据预处理方法进行了研究；其次,针对企业成本控制和客户关系管理问题,本文通过对实际问题的分析和相应数据的处理,对基于企业数据的预测问题进行了研究：最后,为了提高企业管理的有效性和科学性,本文研究了基于企业数据的评价问题。本文的具体研究内容和创新性工作主要有以下几点：1为了有效利用企业数据,针对企业数据的特点,首先本文根据差异最大化思想,构建了相应的降维方法;其次本文研究了基于融合空间思想的线性降维方法。通过相关实验表明,本文构建的数据预处理方法有助于企业数据的有效利用。2本文在深入研究深度学习模型原理的基础上,研究并构建了基于企业数据的交叉混合和顺序混合的深度学习结构,同时为了验证模型的有效性,本文通过收集和整理相关的企业数据,抽取了相应的数据集,并利用本文构建的模型对某电信企业客户忠诚度问题进行了预测研究,通过相应的仿真实验与比较分析表明,本文设计的顺序混合深度学习模型具有相对较好的效果。3本文基于CART分类器的分类误差自动调节Boosting算法相关权重的思想,研究并构建了一种基于企业数据的优化预测模型,同时本文在利用该模型对某电信企业客户离网问题进行预测时,采用了直方图检验和卡方检验相结合的方法对模型输入属性的相关性进行了检验。通过相应的仿真实验表明,本文构建的模型具有相对较好的效果。4本文通过对维修过程的深入分析研究,结合更新过程和更新理论对零部件的更换问题进行本质性描述,构建了基于企业数据的更换率预测方法,在对MRO企业零部件更换率预测过程中,本文通过对企业数据的转换,拟合出零部件更换的概率分布,同时利用最小二乘法对相应的威布尔分布的参数进行估计。将零部件更换率与维修BOM信息结合,可以有效控制MRO企业成本,为精益MRO提供支持。5本文在分析研究企业管理评价问题时,将评价问题与动态系统状态的变化问题相结合,提出了一种基于状态熵的评价模型,并对状态熵计算公式进行了相应的数学推导,通过计算评价目标在系统中的状态熵可以确定评价目标的综合绩效。同时本文将该模型应用于MRO企业供应商评价问题中,从MRO企业角度出发,利用平衡积分卡法构建了相应的评价指标体系,并利用网络分析方法对相应的指标权重进行处理。通过利用企业数据进行相应的仿真实验表明,本文设计的模型具有一定有效性。
[Abstract]:With the development of science and technology, especially the Internet and computer communication technology and the in-depth application, people are inevitably in the ocean, whether in daily life or production and business activities of enterprises, the data are generated. Every hour and moment channel data is constantly increasing, so that people get the data becoming more easily, the cost has been greatly reduced, but also, the increase in the amount of data and data form has led to increasingly complex data by becoming more and more difficult, how to effectively use data processing and data, to play its due role, provide strong support for human life and social development. Has become the focus of discussion in the community. With the data of rising global academia, industry and governments are highly concerned about the related Data technology, data science has gradually become a new scientific research. Because of the rational use of data can not only serve the public, but also can bring huge commercial value for the enterprise, so the data is called "the future of the new oil, the focus of competition in the future will focus on data share, control and use. Although the importance of data has been recognized the world, but because of the uneven social development, many domestic enterprises are still unable to enjoy the data for the benefits of these enterprises is still in the initial stage in the use of data, business data corresponding to the remains of" sleeping "state. Based on the extensive collection and comprehensive analysis of domestic and foreign enterprises on the basis of the research results of data processing and use of the process, the corresponding enterprises conducted in-depth research, combined with the enterprises in the data processing and the Some of the problems existing in the process, determine the research direction of this paper. Firstly, according to the enterprise data complexity and data dimension increasing, enterprise data application problems, the pretreatment methods for relevant data; secondly, according to the enterprise cost control and customer relationship management, this paper according to the actual the analysis of the problems and the corresponding data, the prediction based on data from the company were studied. Finally, in order to improve the scientific and effective management of enterprises, this paper studies the problem of evaluation of enterprise based on data. The specific contents and innovative work are mainly as follows: 1 in order to make effective use of enterprise data. According to the characteristics of enterprise data, firstly, according to the difference of the idea of maximizing, construct the corresponding dimensionality reduction method; secondly this paper researches the integration of spatial thinking based on line Dimensionality reduction method. The experiments show that the data preprocessing method in this paper is helpful to the effective use of.2 based enterprise data learning model principle in-depth research in depth on the basis of the research and the construction of the enterprise data and order mixed cross mixing depth study based on the structure, at the same time in order to verify the validity of the model. In this paper, through the collection and collation of enterprise data, extract the corresponding data set, and the problem of customer loyalty to a telecom enterprise is predicted using the model constructed in this paper, through the simulation experiment, the corresponding analysis and comparison show that the order of mixed deep learning the design of this model has a better effect in this paper.3 Boosting automatic adjustment the algorithm weights of classification error of CART classifier based on the idea, and establish a forecasting model for the optimization of enterprise based on the data, at the same time In this paper, using the model of a telecom enterprise customer churn prediction problem, by using the method of histogram test and chi square test combined with the correlation of the model input attributes are examined. Through the corresponding simulation experiments show that the model has a relatively thorough analysis and Research on the maintenance process through better effect.4 in this paper, combined with the theory of renewal process and update the replacement problem on parts of the nature description, constructs the prediction method of replacement rate based on enterprise data, in the replacement rate of MRO parts enterprises in the process of prediction based on conversion of enterprise data, fitting out the probability distribution of replacement parts, and the corresponding parameters of Weibull distribution are estimated by the least square method. The parts replacement rate and maintenance of BOM information, can effectively control the cost of MRO company, to provide for the lean MRO Support.5 based on the analysis of the problems of evaluation of enterprise management, the problem will be evaluation problem and the dynamic system of the change of state combination, this paper puts forward a model of condition evaluation based on entropy and state entropy formula of mathematical derivation, the comprehensive performance evaluation by the state entropy calculation targets in the system can determine the evaluation at the same time, the problems of the target. The model is applied to the MRO supplier evaluation, starting from the MRO enterprise perspective, using the balanced scorecard method to construct the corresponding evaluation index system, and deal with the corresponding index weight using the network analysis method. Simulation results show that by using the corresponding data of the enterprise, the design of the model has a certain effectiveness.

【学位授予单位】：合肥工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：F272

【相似文献】