Partition of Discrete Numerical Interval Based on Local Outlier Coefficient Wave Method

The partition of discrete numerical interval is the foundation of discrete numerical data mining, and its research is significant in application. A new method for the partition of discrete numerical interval is proposed here. In the method the original samples are preprocessed, and then the local outlier coefficient of each data calculated. Eventually, the discrete numerical interval is partitioned by local outlier coefficient wave (LOCW), which keeps original distribution of data. The partitioned intervals have advantages of data aggregation in same interval and data sparsity between different intervals. The experiments demonstrate the validity of the proposed method.