Research On Application Of The Associative Classification Method In Aluminum Electrolysis

Due to its special physical and chemical properties, aluminum plays an important role in the modern industrial society such as the industries of aerospace, construction and automotive. Now aluminium is mainly produced through electrolysis. How to improve the quantity and quality of aluminum production is the main and urgent problem pressing for solution. The development of the computer softwares has provided a good opportunity for the electrolytic production of aluminum to effectively improve the aluminum yield and quality. Data mining is an important branch of computer science, which focuses on how to extract and make use of the valuable information embedded in a large number of data. Application of data mining in aluminum production has made great progress by now.Classification is one of the main tasks of data mining to predict the category of each unknown object. And the associative classification is an effective method for building classification rules. Some parameters of aluminium production can be predicted in the production process, and provide a useful guide for industrial production. In this paper, we studied the associative classification algorithm and its application in the aluminium production.Firstly, the present frequent itemset mining algorithms have some disadvantages, for example, Apriori has a huge I/O overhead, FP-growth algorithm consumes huge memory. Thus, based on the structural features of the FP-tree. we proposed a new FP-tree based frequent itemset mining algorithm named dynamic paine. This dynamic prune algorithm not only kept the high efficiency of FP-growth algorithm, but also solved the problem of high memory consumption of the FP-growth algorithm. In the meantime, it accelerated excavation by using concurrent mining. Therefore, this dynamic prune algorithm is better than the Apriori and FP-growth algorithms in terms of timeliness and scalability.Secondly, the traditional associative classification methods used the Apriori algorithm or FP-growth algorithm to generate the class association rules, so they either cause large I/O overhead, or take up a huge amount of memory. Here we designed an improved method for class association rule generation based on cutting FP-tree method, which is more feasible and efficient than traditional methods.Finally, based on the improved associative classification method mentioned above, combined with aluminum electrolytic production parameters, we designed and used the associative classification system in the electrolyzer. This system included all steps of associative classification method. Each step gave the differences between different algorithms. At last, we calculated the prediction accuracy by experiments, and provided support for the aluminum electrolytic production decisions.