冷轧带钢表面缺陷匹配信息推荐算法设计与实现

发布时间：2018-02-08 15:24

本文关键词： 冷轧带钢表面缺陷信息内容推荐协同推荐混合加权推荐　出处：《武汉科技大学》2016年硕士论文　论文类型：学位论文

【摘要】：目前在冷轧带钢表面缺陷处理过程中,由于缺陷信息复杂多变且大都采用人工分析缺陷产生原因的方式,导致相关用户实施整改措施滞后,产品缺陷影响持续扩大。针对这一问题,本文结合当代互联网广泛应用的信息智能推荐方法,对冷轧带钢表面缺陷匹配信息推荐算法进行了研究。根据带钢相关的缺陷信息、用户信息和经验信息的特点,结合内容过滤和协同过滤的优势,采用混合加权的方法,提出一种适用于带钢表面缺陷的混合推荐算法。通过带钢表面缺陷实例数据的分析验证表明,该混合推荐算法对比单一算法具有更高的准确性和用户接受度。本文主要的研究工作如下:(1)对冷轧带钢表面缺陷信息、用户信息和经验信息进行分析和整理,全面总结信息数据的特点,在基于专家经验的基础上建立经验信息映射关系数据表,并根据冷轧带钢表面缺陷信息与用户信息的关联特点,确立打分评价体系。(2)在满足冷轧带钢表面缺陷匹配信息推荐系统对推荐算法的实际需求下,结合基于内容过滤算法和协同过滤算法各自的优点,采用混合加权的方法,设计一种应用于冷轧带钢表面缺陷匹配信息的混合加权推荐算法,其中采用夹角余弦相似度方法作为文本推荐算法,基于贝叶斯网络模型法作为协同过滤推荐的算法。(3)根据冷轧带钢产线实际情况,对其表面缺陷信息进行推荐算法的验证研究。案例中确定混合加权推荐算法的权重系数等值,分别计算得出基于内容推荐算法、基于协同推荐算法和混合加权推荐算法的推荐结果,并从推荐结果、推荐精度、用户接受度等方面对算法进行分析比较。最后设计冷轧带钢表面缺陷匹配信息推荐算法的三个主要软件功能模块:缺陷匹配信息推荐模块、打分数据库模块和文本内容信息库模块,并通过Visual Studio 2013软件开发平台实现冷轧带钢表面缺陷匹配信息推荐算法的功能。
[Abstract]:At present, in the process of surface defect treatment of cold-rolled strip, due to the complex and changeable defect information and the manual analysis of the causes of defects, the relevant users are lagging behind in the implementation of corrective measures. The influence of product defects continues to expand. In view of this problem, combining with the information intelligent recommendation method widely used in contemporary Internet, this paper studies the recommendation algorithm of surface defect matching information for cold-rolled strip steel, and according to the defect information related to strip steel, The characteristics of user information and experience information, combined with the advantages of content filtering and collaborative filtering, are combined with the method of mixed weighting. A hybrid recommendation algorithm for surface defects of steel strip is proposed. Compared with the single algorithm, the hybrid recommendation algorithm has higher accuracy and user acceptance. The main research work of this paper is as follows: 1) the surface defect information, user information and experience information of cold rolled strip are analyzed and sorted out. The characteristics of information data are summarized in an all-round way. Based on the expert experience, the relational data table of empirical information mapping is established, and according to the characteristics of the correlation between the surface defect information of cold rolled strip and user information, In order to meet the actual requirement of the recommendation system for the surface defect matching information of cold-rolled strip, combining the advantages of the content-based filtering algorithm and the cooperative filtering algorithm, the mixed weighted method is adopted. A hybrid weighted recommendation algorithm applied to the surface defect matching information of cold-rolled strip is designed, in which the angle cosine similarity method is used as the text recommendation algorithm. Based on Bayesian network model method as a collaborative filtering recommendation algorithm, according to the actual situation of cold-rolled strip production line, the surface defect information of the recommendation algorithm is verified and studied. In the case, the weight coefficient of the hybrid weighted recommendation algorithm is determined. The recommendation results of content-based recommendation algorithm, collaborative recommendation algorithm and hybrid weighted recommendation algorithm are calculated respectively. Finally, three main software function modules of the recommendation algorithm for surface defect matching information of cold rolled strip are designed: defect matching information recommendation module, The database module and the text content information database module are graded, and the function of recommending information algorithm for surface defect matching of cold rolled strip is realized by Visual Studio 2013 software development platform.
【学位授予单位】：武汉科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TG335.56;TP391.3

【相似文献】