Approach for recognition of true and false specific sample points

In the process of data mining, a major obstacle of using mathematical analysis to study the patterns and trends hidden in the data is the specific sample points existed in large-scale data sets. According to the ratio of specific sample points to the sample size, taking into account other factors at the same time, specific sample points may be divided into true and false specific sample points. For the first time, the paper proposes the discriminant formula for recognizing true and false specific sample points based on the first m principal components, which is also base on the re-definition of discriminant formula for the first 3 principal components. The construction principle of critical value formula is also analyzed. These concepts, formulas and conclusions have important reference values on eliminating the sample points interfered with random factors, a reasonable selection of index set and refining models.