Recognition Method of Abnormal Data Based on Interval Estimation

The emergence of substantial abnormal data in network data acquisition system is due to the cognitive level of users mostly. The abnormal data not only cause the waste of storage and the operation resources, but also seriously affect the results of statistical analysis, How to identify the massive abnormal data is the premise to eliminate the abnormal data. Accordingly a detection method of abnormal data based on confidence interval, which is aimed to filter the abnormal data from the group, was proposed. Compared with the traditional method of simple random sampling, the sample space is selected from the overall sample by the above-mentioned method has more credibility. And different from the commonly used detection method which is aimed at abnormal point, this method can batch test the unexpected data. The experimental results suggest that this method can effectively filter out the abnormal data particularly in the cases where the data size is huge and the data deviation is not obvious. And on the basis of above-mentioned method, the detection result has little difference with the actual situation and the accuracy is over 90%.