Short-term traffic volume forecasting: A k-nearest neighbor approach enhanced by constrained linearly sewing principle component algorithm.

Highlights: [•] A novel algorithm is proposed to forecast short-term traffic volume. [•] Overlapping is avoided in selecting k-nearest neighbors. [•] A linearly sewing principle component algorithm is developed. [•] The new algorithm outperformed the competing algorithms in most cases. [Copyright &y& Elsevier]