Abstract:

Missing data treatment is an important content of data cleaning. A classification detection method for uncompleted records is proposed. The uncompleted record is defined and records are classified as four classes, including completed records, uncompleted and unmodifying records, uncompleted and modifying records, uncompleted and deleting records. A classifying flow with hiberarchy is given. The binary expression of a record is defined. The standard binary expression sets of each class are created according to uncompleted record samples. Priority of standard binary expressions is determined by appearance times in samples. Some specific binary expressions are merged using formulas. Classification detection of records is implemented by bit operation.Binary expression sets are perfected step by step through dealing unseen binary expressions. The next processing of uncompleted records could be confirmed by their binary expressions. The effectiveness of the proposed method is validated by an instance.