1、基于HIFAHP方法的电力客户欠费风险评价与预测模型Power customer arrears risk assessment and prediction model based on HIFAHP method摘要电费回收管理一直是供电企业的工作重点，供电企业采用先用电后交费的规则来管理，因而存在电费回收周期长、催费措施落后等问题，这些问题长期困扰着供电企业。此外，缺少有效的欠费回收风险分析手段，未建立差异化风险防控策略。目前只能通过人工或系统的分析工具单一的识别电费回收风险，一方面未根据客户的基本情况、行业特征、行为特点等信息进行全方位融合分析，无法精确甄别高风险客户，另一方面由于风险
2、防范的时间点相对滞后，没有形成事前的标准化预警机制，导致难以及时制定针对性的防控措施，加大了企业的运营风险。鉴于此，为了有效提高电力企业的风险防控能力，降低企业经营风险，高效准确的预测欠费风险，基于用电客户的海量历史数据信息，借助大数据分析技术，建立集评价与预测一体化的欠费管理方法。高效准确的识别欠费高风险客户，从而帮助供电企业制定催费措施，提高回收率。论文基于目前供电企业电费回收管理情况和相关行业风险评价理论，将电力客户分为两大类，根据两类客户特点分别构建电力客户欠费风险评价指标体系。针对建立的评价指标体系，提出了基于犹豫直觉模糊层次分析法。依据层次分析法构造犹豫直觉判断矩阵，同时运用位置权
3、重和正态分布赋权法确定属性权重，从而计算犹豫直觉模糊矩阵的得分函数，对电力客户欠费风险进行评价。在此评价基础上，基于国网大数据平台软件资源，并根据两类电力客户的特点，相应的运用Logistic回归算法和决策树算法对两类电力客户欠费风险进行预测。最后以四家电力客户为实际应用背景，通过3.2节描述的相关处理指标评价，对四家电力客户建立多属性评价矩阵，并计算出变换后的决策矩阵得分函数，从而判断出高风险电力用户。同时分别对建立的两类电力客户欠费风险预测模型进行实证研究，其结果均达到了预测效果。说明了该指标体系及其预测模型对于电力客户欠费风险评价具有良好的效果，实现了对电力客户欠费风险进行科学、客观、公
4、正评价与预测的目的。关键词：电力客户欠费风险；犹豫直觉模糊层次分析法；Logistic回归模型；决策树算法AbstractThe tariff recovery management has been the focus of the power supply enterprises. For a long time, the power supply enterprises have always adopted the market rule that customers can use electricity before buy it. While the recovery tarif
5、f cycle is always long and the measures to urge the fee are backward, then the tariff recovery has been a significant problem in the power supply enterprises.In addition, there is no effective risk analysis method for arrears, and the power supply enterprises havent establish differentiation risk co
6、ntrol strategy.At present, only manual or systematic analysis tools can be used to identify the risk of the tariff recovery. On one hand, we cant carry on the comprehensive analysis without considering the basic information, industry characteristics and behavior characteristics of customers. On the
7、other hand, because of the backward of the practice point of risk prevention. We cant form early standardizedwarning mechanism which will make it difficult to formulate differentiated control measures in time and increases the operational risk of the enterprises.In view of this, we should build a se
8、t of management method to estimate and predict arrears by using big data analysis technology based on the massive historical data information of customers. Then the power supply enterprises will improve the risk prevention and control ability, reduce the risk of enterprise operation and predict the
9、risk of arrears efficiently and accurately.This paper will divide the electric power customers into two categories based on the situation of the tariff recovery management in the power supply and the industry risk assessment theory to construct the risk evaluation index system of power customers arr
10、earsrespectively.To study the system, a kind of analytic hierarchy process based on hesitant intuitionistic fuzzy sets was proposed. We can construct the hesitate intuitionistic judgment matrix depending on the importance of the property and determine thepropertyweights according to the weights of p
11、ositions by normal distribution weighting method. Then we will get the score functions of hesitant intuitionistic fuzzy matrix to evaluate the risk of electricity customer arrears.Logistic regression arithmetic and decision tree algorithm will be applied in evaluating the arrears of the two classes
12、of power customers according to their characteristicsbased on the national network big data platform software resources.This paper takes four electric customers as the practical application background. The multi-attribute evaluation matrix will be established for the customers by the evaluation of r
13、elevant processing indexes described in section 3.2. By a series of transformation we can get the score functions of decision matrix, and we will identify high-riskpower customers.At the same time, the two prediction models have been empiricallyresearched. The results achieved the prediction effect
14、very well. Therefore, the index system and prediction models have good effect on the risk assessment of power customers arrearsto evaluate and predict the risk of electricity customer arrears scientifically, objectively and justly.Key Words：Electricity customer arrears risk；Hesitant intuitionistic f
15、uzzy analytic hierarchy process；Logistic regression model；Decision tree algorithm目 录摘要IAbstractII1 绪论11.1 研究意义11.2 国内外研究现状21.2.1 决策分析研究现状21.2.2 电力客户欠费风险研究现状31.3 研究内容41.4 文章结构42 预备知识62.1 直觉模糊集理论62.2 犹豫模糊集理论72.3 犹豫直觉模糊集72.4 Logistic回归模型102.5 决策树算法113 电力客户欠费风险评价123.1 电力客户欠费风险评价指标体系123.1.1 客户的分类133.1.2
16、电力客户欠费风险评价指标体系的建立143.2 基于犹豫直觉模糊层次分析法在电力客户欠费风险评价中的应用183.2.1 数据预处理193.2.2 犹豫直觉判断矩阵的建立203.2.3 一致性检验及修正213.2.4 指标权重的确定223.2.5 综合评估264 电力客户欠费风险预测模型构建与算法设计274.1 基于Logistic回归模型的欠费风险预测模型274.1.1 建立Logistic回归预测模型274.1.2 使用Logistic回归模型应注意的问题294.2 基于决策树算法的欠费预测模型304.2.1 数据标准化及数据分区304.2.2 模型验证及测试314.3 结论325 案例分析345.1 电力客户欠费风险评价345.2 Lo