Abstract: The author proposed a characteristic trend reasoning algorithm for largescale data. Firstly, Hash function was used to extract largescale data samples, Pam clustering algorithm and parallel Kmeans clustering algorithm were used to
 cluster largescale data samples. After obtaining the best clustering results, the dynamic characteristics of large data clustering were extracted. Secondly, a reasoning algorithm based on characteristic trend rules was used to construct a trend rule reasoning model for largescale data characteristics, and a trend rule algorithm was designed by the method of cumulative trend rule, which could infer the trends of largescale data characteristics, and solved the problem of large errors of reasoning results. The experimental results show that the average accuracy of the proposed algorithm for large-scale data characteristic trend reasoning is 9810%, the growth rate of reasoning speed is 50%, and the maximum average reasoning timeconsuming is only 11425 s, which can quickly and accurately complete data characteristic trend reasoning.

Key words: largescale data, characteristics, trend, reasoning, dynamic characteristics, cumulative trend rule