Within the scope of recommendation systems, content-based and collaborative filtering recommendations encounter difficulties when dealing with the cold start, particularly when considering the temporal evolution of item popularity. These algorithms tend to recommend highly popular items, neglecting the dynamic nature of item hotness over time. In the express industry, users independently choose items for delivery, and many of these items exhibit seasonal patterns. Traditional recommendation methods may struggle to effectively capture the current popular items.To address the cold start issue in express item recommendation, this paper introduces an adaptive hot items ranking value (ADHV) algorithm. This algorithm accounts for temporal changes in express item volume and overall popularity. It adjusts the base hot value of each item using Bayesian ideas to determine its true hotness.This paper also proposes a metric called Hot Item Weighted Evaluation (HWE) to address the issue of the inability to quantitatively evaluate cold start recommendations. The HWE metric combines Newton’s cooling method with Spearman’s idea. The experimental results compare different hot ranking recommendations algorithms, and the HWE metric confirms the effectiveness and robustness of the ADHV algorithm.