工业大数据分析技术与轮胎销售数据预测

Data analysis of industrial big data and sales forecast of tyre industry

工业大数据是在工业领域信息化应用中所产生的海量数据,作为决策问题服务的大数据集、大数据技术和大数据应用的总称。首先分析工业大数据4V特性与工业数据的特有特征,以及工业大数据来源;从多源异构工业数据集成与数据融合方法、工业大数据计算架构、大数据带来的信息安全等三方面论述工业大数据面临的挑战与潜在价值。探讨了工业大数据分析与挖掘方法,提出了工业大数据平台的计算架构与大数据处理平台,构建轮胎企业大数据资源中心、大数据分析与决策应用系统。从销售数据分析和宏观数据趋势两个层面进行轮胎销售大数据分析与预测。采用多个不同领域的销售数据源来解决销售预测历史数据特征空间稀疏的问题,使用LASSO(The Least Absolute Shrinkage and Selectionator Operator)方法的多任务学习方法来解决高维样本空间的缺点,实验数据验证能够提升轮胎销售预测的准确率。

Industrial big data,the huge amount of data produced in the informatization process of industry field,is a collection of big data set,big data technology and big data application targeted to serve decision-making.Firstly,the4V characteristics of industrial big data and the special features of industrial data are analyzed,along with the source of industrial big data.After that,challenges and potential value of industrial big data are expounded from three perspectives:industrial data integration and fusion method for multi-source and heterogeneous data,computing architecture of industrial big data supporting real-time modeling,information security issues brought by big data.Industrial big data analysis and mining methods are discussed.The computing architecture and processing platform of industrial big data are proposed for the requirements of enterprise industrial big data application.This paper constructs an enterprise-level big data resource center and a big data analysis and decision system based on the big data application requirements in the tire industry.By integrating and utilizing multiple data sets concerning sales,it executes tire sales big data analysis and forecasts from two levels:sales data analysis and macro data trends.Multiple sales data origins from different fields are used to solve theproblem of sparse sales forecast history data feature space.Multi-task learning method based on LASSO(The Least Absolute Shrinkage and Selectionator Operator)is used to overcome the shortcomings of high-dimensional sample space.Data from experiments prove that the accuracy of tire sales forecast is improved.