Group 1 - New home sales are slowing down, while the used housing market remains resilient, and passenger car sales continue to show strong performance [9][10][19] - In the week of March 10-16, the transaction area of commercial housing in 30 major cities decreased by 26.4% year-on-year, with first, second, and third-tier cities showing declines of 6.0%, 38.2%, and 33.0% respectively [10][16] - The retail sales of passenger cars in February increased by 26.0% year-on-year, driven by the resumption of subsidy policies and stable price expectations from car manufacturers [19][20] Group 2 - Construction demand remains weak, with manufacturing activity showing a slight increase [11][39] - The price of rebar and hot-rolled coils changed by -0.6% and +1.2% week-on-week, respectively, indicating a mixed performance in resource prices [11][26] - The national cement price index increased by 3.1% week-on-week, supported by industry self-discipline and limited production [11][35] Group 3 - Long-distance travel has increased, while the loading capacity of departing ships has decreased year-on-year [12][24] - The passenger volume of subways in major cities increased by 0.9% week-on-week, while the volume of express delivery increased by 20.5% year-on-year [12][24] - The SCFI/BDI index showed a week-on-week change of -8.1% and +19.2%, indicating fluctuations in export activity [12][24] Group 4 - The coal price at Qinhuangdao Port was reported at 681 yuan/ton, with a week-on-week decrease of 1.3% [43][46] - The prices of copper and aluminum increased by 2.8% and 0.7% week-on-week, respectively, driven by expectations of interest rate cuts [47][49] - The total inventory of copper and aluminum decreased by 35,000 tons and 25,000 tons week-on-week, respectively [49][51]

金融工程 | 金工定期报告 金融工程 证券研究报告 在当前市场中，分析师报告数量众多，以行业报告为例，每周通常有超过 500 篇的报告。而阅读这些报告通常要花费大量的时间和成本。针对这一 需求，我们利用 DeepSeek-V3 大模型的总结能力对分析师行业报告进行智 能总结和整合，提炼出核心观点和关键信息。 利用大模型衡量景气度 行业研究报告通常提供分析师对行业趋势的深入分析与总体评价，但其应 用性较弱，主要有两个原因。首先，行业评级是行业研报中的标准化输出， 但其并非连续指标，通常仅分为三类，缺乏足够的区分度。因此，分析师 细微的态度和用词变化可能不会导致行业评级的调整。其次，不同证券公 司采用的行业分类标准并不一致，这些差异使得行业比较变得困难。 大模型普及之前，并不容易解决这些问题，而大模型的应用可能为此提供 更有效的支持。为了深入挖掘行业研报的信息，我们对其进行了进一步的 整理与标准化。我们构造了一个能够提取行业研报所涉具体中信一级、二 级行业名称以及对应行业景气度的提示词，并使用 DeepSeek-V3 模型，将 研报摘要作为输入文本得到该研报所属行业、景气度等指标结果。 最新行业研报文本景气度 我 ...