Geodesic-based Kernelizing K Nearest Neighbor Conformal Predictor

摘要

An improved algorithm was proposed to overcome the shortcomings of the existing conformal predictor algorithm that was not suitable for linearly inseparable data and did not use the information of distant points. First, the geodesic-distance was introduced into the new algorithm to reflect the implied geometry of the data, thus, the algorithm could take advantage of the information of the distant points, and then the kernelizing nonconformity predictive function was designed by using the RBF kernel, which make the algorithm robustness and better to process nearly inseparable data. The results of the experiments on UCI data sets showed that the improved algorithm could obtain better performance than the existing conformal predictor algorithm on classification.