Multi-relational Association Rules And Its Application In HRM

Multi-relational association rule mining is a branch of multi-relational data mining, which by analyzing the data between the relational tables to find the association rules which existing in a single table and between attribute values of multiple tables.The implementation of classical association rule mining is based on the relational database data with a single table. However in practice, the data are mostly stored in relational database spread over multi-tables, if applied the traditional association rule mining techniques directly in the multi-tables, will cause the performance degradation, statistical deviation, loss of information, data redundancy and many other issues. The multi-relational association rule mining is a process, by anylizing the data of multi-tables in relational database to find the association rules with a single table and the attribute values in multiple tables. This technology can not only shorten the process of knowledge discovery, but also can improve the algorithm efficiency and accuracy.Based on the anaylysis of the technology in multi-association rules mining, this paper proposed an improved multi-relational association rules mining algorithm, which has been applied to the Human Resources Manage System of an enterprise. The main research works are concluded as follows:(1) Analyze the association rule mining algorithm of the single-table and the basic principle of ILP (Inductive Logic Programming) technology in multi-relational association rules mining, discussed and studied classical algorithms which based on inductive logic programming, such as WARMR algorithm, FARMER algorithm and so on.(2) Analyze and conclude the advantages and shortcomings of the multi- relational association rules mining which based on ILP technology: ILP algorithm has resolved the problem of statistical skew preferable, but it greatly dependents on theθcontains and the key atom, so each procedure can only find the association rules bound up with the key atom. All the association rules can be mined only by exchanging the key atoms. Therefore, the algorithms of multi-relational association rules mining which based on ILP technologies are difficult to apply to the project of actual data mining.(3) Proposed an improved multi-relational association mining algorithm MID_ CrossMine. CrossMine algorithm used the ID tuple propagation technology to achieve the virtual connection and then the data mining of multi- relational tables. However, in the process of the tuple ID propagation, it needs to set classification label necessary, and selection of the classification label has a strong empirical. The algorithm MID_CrossMine which proposed in this paper took the result of 1-itemsets as the references of classification and combined with MTPA algorithm to solve the defects of CrossMine and enhance the efficiency of mining.(4) Selected a human resources database of an enterprise as mining background and applied the MID_CrossMine algorithm to the analysis of human resources management system.In it analyzed the staff composition, performance appraisal, personnel mobility and other aspects. The result of relational association rules can help enterprises to establish a good early-warning mechanism and provide references of personnel management system to the managers.