Clustering Method Study on High-Dimensional Trading Data

Existing clustering algorithms are not designed specially for the features of trading data s and most clustering analyses lack scalability for large-scale transactions. Therefore, a rapid and scalable clustering algorithm using little space is proposed by us, to effectively process high-dimensional trading data without setting parameters manually. The improved method introduces weighted coverage density as similarity metrics of data. On this basis, the clustering criterion function is established for clustering analysis. We assume further implementation is to find association rules in clustering rules. Then two transaction-oriented evaluation measures for clustering quality are put forward. The large item size ratio is based on the concept of big data, which is used to measure the percentage in clustering; the average pair-clusters merging index is adopted to indicate the difference among clustering results with coverage density. The experimental results of artificial data and real data sets have shown that the improved method for clustering analysis can generate high-grade clustering results on most of the experimental data sets, compared to traditional algorithms