维州行人事故：热点与道路条件分析Victorian Pedestrian Crashes: Hotspots and Road Conditions
事故集中在哪里？这些地方的事故，在天气、路况等条件上与全州有什么不同？Where do crashes concentrate, and how do their recorded conditions differ from Victoria overall?
- 项目背景：课程作业要求为澳洲维州交通部门做决策看板，判断应优先在哪些区域改善行人安全。
- 工作内容：整合 4 张官方表、1.7 万起事故；圈选区域可按天气、光线、路况与全州对比事故占比，注明事故多可能只因人多。
- 项目结果：看板显示 65 岁以上行人占事故 20.4%，却占死亡事故 41.8%；作业 84.3/100（班均 70.0），获最高档 H1。
- Coursework: a decision dashboard for Transport Victoria on where to improve pedestrian safety first.
- Joined 4 official tables (17,510 pedestrian crashes) into a map filterable by year, time of day and age; select any area to compare its share of crashes under weather, light and road conditions with the whole state; the page notes that more crashes may just mean more people.
- Scored 84.3/100, an H1 (top band; class mean 70.0). The dashboard shows people 65+ are 20.4% of crashes but 41.8% of fatal ones.
- 数据Data
- 维州交通规划部 4 张官方表（2012–2025 年的事故，CC BY 4.0）；墨尔本市行人流量4 official tables from Victoria's Department of Transport and Planning (crashes 2012–2025, CC BY 4.0); City of Melbourne footfall
- 视图Views
- 热点、事故点、人流三种地图Three maps: hotspots, crashes, footfall
- 筛选Filters
- 年份、严重程度、时段、行人年龄Year, severity, time of day, pedestrian age
- 作业要求The brief
- 为明确的决策者设计：帮维州交通部门调查行人事故并采取行动Design for a named decision-maker: help Transport Victoria investigate pedestrian crashes and act on them
打开浏览器就能用的网页版A web version that runs in the browser
另做了一个网页版，用的是同一份公开数据：17,510 起行人事故和 84 个行人流量传感器。热点、事故点、人流三种地图和筛选都在。I also built a web version on the same public data: 17,510 pedestrian crashes and 84 footfall sensors, with the hotspot, crash and footfall maps and the filters.
提交版的六种画面Six views of the submitted version
看板是 Tableau 文件，这里放提交版（v36）的六种画面，交作业前逐一检查过。点图看大图。The dashboard is a Tableau workbook, so these are the six states of the submitted version (v36) that were checked one by one before submission. Click to enlarge.
我做的三个取舍Three trade-offs I made
从区域差异中找调查线索Find leads in local differences
圈选事故集中区域，在同样筛选条件下，与全州比较天气、路面、光线、限速和道路形态下的事故占比。用百分点差呈现当地的突出特征，帮助确定值得进一步调查的道路情境。Select a crash cluster and compare its shares under weather, surface, lighting, speed-limit and road-geometry conditions with Victoria under the same filters. Percentage-point gaps highlight local patterns worth investigating.
把局限写在图上Put the limits on the dashboard
条件占比的差异提供调查线索，不能据此认定某个条件导致事故。事故数也不等于按出行量调整后的风险；人流传感器只作背景，没有与事故记录关联。Differences in recorded conditions provide leads, without establishing that a condition caused crashes. Counts are not exposure-adjusted risk; footfall sensors provide context and are not joined to crash records.
一个看板，三种视图One dashboard, three ways to look
热点看哪里多，事故点看具体哪一起，人流看哪里人多；三种视图共用同一组筛选。Hotspots show where crashes cluster, crash dots show individual cases, footfall shows where people are; all three share one set of filters.
H1（最高档），84.3/100，班均 70.0（291 人）84.3/100, an H1 (top band); class mean 70.0
老师评语的开头：“outstanding technical and conceptual work”。The marker's feedback opens with “outstanding technical and conceptual work”.