Abstract:
The paper proposed a basic hypothesis for deducing density function and derived density function. It achieved density regional division through the density function, gave a specific estimating approach for computing tag value of untagged label within the same density range, and described specific implementation steps of semi-supervised regression algorithm based on density distribution. The algorithm achieves mark of untagged points, and can reduce estimating error of the tag value and improve estimation accuracy.