Kernel regression is widely used in biology and economy, because it is more adaptable to complex laws than linear regression, and it has better interpretability than many methods in deep learning. In highdimensional area, l1-norm penalization is a common method for variable selection, which may be derived from the excellent performance of the lasso algorithm. Although it seems natural to generalize from consistency in variable selection to consistency in kernel selection, there are still many details that need to be taken seriously, e.g. the lower eigenvalue condition. The consistence condition of kernel selection in l1-norm regular linear kernel regression is given, including the prediction error. In simulation study, the consistency of different levels of λn and dimension of features is carefully checked. Finally, the kernel selection method was applied to high-risk area exploration of Covid-19, with the dataset provided by the US Centers for Disease Control and Prevention(CDC). Declarations of interest:The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.