The Research On Clustering And Classification Algorithm Data Mining And Its Application In Aluminum Data Analysis

As the computer industry in the production of aluminum electrolysis to promote the application of various plants in the production process are involved in the use of a computer monitoring system to achieve automatic control of the cell. Various slot monitoring system status data was collected automatically, the electrolytic production industry has accumulated a lot of historical data. However, the data sharing and data integration of the existing system are low, and it is only for simple data entry, query, statistics and other matters of process, but it can not find these huge volumes of data contained in the production and management of enterprises has important rules and laws of the guiding role. The Policy-makers urgent need extracted valuable information and knowledge from the massive data to electrolyze for the management, and increase production efficiency. For aluminum electrolysis cell status data analysis, the main contents and contribution can be followings: 1.In the data preprocessing of data, Use Interpolation methods to fill hole, according to the inertia of system running, analysis data of hole before and after to get the similar forecast result as a value of the hole, for the realization of data mining data to support.2.A algorithm of connected components cluster based on gray association degree (Gry-CC-CTL) is proposed for aluminum electrolysis data. First the algorithm analyzed the relationship between the main attributes and the main attribute, and calculated the gray weight, then applied them in distance calculation of connected components clustering. The actual production data as an example of the experimental results showed that the algorithm has effect. Finally, a simple slot status cluster analysis system is designed and realized.3.On the basis of historical data cluster analysis of electrolytic, consider the impact of historical data on the current operation, A ReliefF weighted classification algorithm of k-neighbors distance recognition is proposed for Dynamic classification of real-time data. First ReliefF method calculated the inspect attributes of the contribution of classification to embody all the attributes of the classification level of contribution, calculated contribution weight, then applied it to the calculation of the distance, set the threshold value method to determine the status of the trough anomaly; Experimental results show that the algorithm to slot the status of data classification and the accuracy of the results have better performance. Finally, a simple classification system slot status is designed and implemented.