基于时间序列算法的网银交易量预测

发布时间：2018-01-06 16:40

本文关键词：基于时间序列算法的网银交易量预测　出处：《东华大学》2016年硕士论文　论文类型：学位论文

【摘要】：在“数据爆炸”的大数据时代,巨大的数据量,价值密度低的数据特点促使人们学习从庞大的数据转换成有用的信息和知识的技能。数据挖掘是大数据时代的产物,它是从庞大数据中发现潜在的有用信息的过程。时间序列分析是大数据分析中很常见的一个部分,时间序列预测又占据了十分重要的地位,它是指根据历史数据,利用科学的方法和技术进行合理的分析,发现其中的规律,最终实现对事物发展趋势的评估。时间序列分析的挖掘与预测具有非常重要的现实意义,被广泛运用在宏观经济、企业管理等方方面面,特别是对于金融行业的发展和金融规律的研究。本文实践基础是基于浦发银行网银交易量预测项目。网银越来越多的进入到人们生活中,网银交易成为银行一大主要业务。每日网银交易量形成一个时间序列,所以我们试图建立合适的模型,让经营者对未来的数据进行预测并对自己的业务作出相应调整。本文首先分析国内外学者对时间序列的特性、分析方法以及预测进行的相关研究,并将其最终运用到实际项目中。数据由浦发银行提供,通过预测分析的基本方法:回归法、移动平均法来分析浦发银行网银交易量历史数据,从而预测2014年5月后每日网银交易量。取得预测结果之后,通过预测结果即数据曲线的拟合程度来判断该系统的合理性和可信性,最终搭建完成时间序列预测系统。本文完成工作包括:(1)分析不同时间序列模型的优缺点,分析时间序列、数据挖掘的发展,并阐述数据处理的一般方法进行数据预处理,增强数据预测能力,优化数据结构和质量,达成最终预测结果准确率提升的目的。(2)分析阐述算法的用法,对时间序列经典算法进行综合比较并确定合适此项目的时间序列预测算法。最终确定本项目利用Microsoft时序算法,这是使用ARIMA与ARTXP相结合的时间序列预测算法。(3)实现时间序列预测平台搭建与完善,实现网银交易量分析,改变系统参数配置,优化数据处理等方式,完善系统的使用性。
[Abstract]:In the era of big data explosion, a huge amount of data, the data characteristics of low value density prompted the conversion from large data into useful information and knowledge learning. Data mining is the product of the era of big data, it is found that the process of potentially useful information from huge data in time series analysis. Is a very common big data analysis, time series prediction and occupies a very important position, it is based on historical data, a reasonable analysis of the use of scientific methods and technology, find the rules, the final assessment of the development trend of things. Has very important practical significance to mining and prediction of time series analysis, is widely used in macro economy, enterprise management and other aspects, especially the research for the development of Finance and financial industry. This paper is based on practice Prediction of Shanghai Pudong Development Bank online banking transaction volume of online banking projects. More and more into people's life, become a major online banking transaction banking business. The formation of a time series of daily online banking transaction volume, so we try to establish an appropriate model, so that operators for the future data to predict and make corresponding adjustments to their own business in this paper. First analyzes the domestic and foreign scholars on the time series characteristics, relevant research and analysis and forecast method, and finally applied to the practical project. The data provided by the Shanghai Pudong Development Bank, through the basic methods of predictive analysis: regression analysis of Shanghai Pudong Development bank online banking transaction amount of historical data to predict the moving average daily after May 2014 online banking transaction volume. The prediction results, the prediction results that the fitting degree of data curve to determine the system's rationality and credibility, and ultimately To build a complete time series forecasting system. This work includes: (1) analyze the advantages and disadvantages of different time series models, time series analysis, the development of data mining, and describes the general method of data processing for data preprocessing, enhanced data prediction ability, optimizing the data structure and quality, to reach a final prediction accuracy rate increase. (2) analysis algorithm usage, a comprehensive comparison of the classical time series algorithm and determine the appropriate time series prediction algorithm. This project to determine the final project using the Microsoft time series algorithm, this is the time series using ARIMA combined with ARTXP prediction algorithm. (3) time series forecasting platform to build and perfect analysis and implementation of online banking, trading volume, change the system configuration parameters, optimization of data processing, improve the use of the system.

【学位授予单位】：东华大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.13;O211.61

【参考文献】