关键词: 大数据, 工业大数据, 工业大数据计算架构, 销售预测

Abstract: 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, the 4V 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 the problem 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.

Key words: big data, industrial big data, computing architecture of industrial big data, sales forecast