| --- |
| title: Traffic Exception Explorer |
| emoji: 🚦 |
| colorFrom: blue |
| colorTo: yellow |
| sdk: static |
| pinned: false |
| --- |
| |
| # Traffic Exception Explorer |
|
|
| An interactive prototype for locating and inspecting recurring high-occupancy traffic records. |
|
|
| ## Data contract |
|
|
| **Purpose.** Help a prototype user identify, compare, and inspect recurring high-occupancy values by road segment and interval. |
|
|
| **Community and authority.** The proposed audience is Kunshan commuters, families travelling near schools, and planning staff. Their need is a hypothesis that needs co-design validation; this Space has no authority to make claims about Kunshan conditions. |
|
|
| **Action and target.** The reader filters, identifies, compares, and inspects repeated high values in a source-series-by-hour matrix. |
|
|
| **Data boundary.** The interface reads `traffic_sample.json`, a small derived sample from the Traffic subset of [Duyu/Time-Series-Forecasting-Benchmark-Datasets](https://huggingface.co/datasets/Duyu/Time-Series-Forecasting-Benchmark-Datasets). The source file is `Traffic.csv`, repository revision `3341455`, accessed 2026-09-06. The sample preserves six source-series identifiers, eight consecutive hourly timestamps beginning 2016-07-01 02:00:00, and the original occupancy values converted from a 0-1 rate to percent. It is historical Bay Area freeway data and is not evidence of current Kunshan traffic conditions. |
|
|
| **Supported claim.** A selected displayed record is high relative to the stated threshold and its segment baseline. |
|
|
| **Prohibited inference.** The view cannot establish a real Kunshan traffic condition, congestion cause, accident, roadwork, traveller behaviour, equity effect, or policy outcome. |
|
|
| **Operations.** The local sample records its source file, revision, access date, and extraction rule. Corrected samples must retain this provenance. Withdraw or relabel the view if source provenance or reuse conditions cannot be verified. |
|
|
| **Success test.** A tester can use a filter, locate a repeated pattern, explain why a cell was flagged, find the displayed raw record, and name one limitation. |
|
|
| ## Interface and interactions |
|
|
| - The **heatmap** is the first analytical view: rows are retained source series and columns are historical hourly intervals. |
| - Filters change the displayed values and the key finding. The threshold determines what counts as high occupancy. |
| - Every heatmap cell is keyboard-focusable. Selecting a cell updates the detail panel and linked trend chart. |
| - All records in this derived sample are labelled `observed`. The interface retains an explicit data-status control so a later documented sample can distinguish unavailable or transformed values without treating them as normal readings. |
| - **Simulate source error** demonstrates the loading/failure state. It does not erase the source and limitation information. |
| - **Analysis trail** records only the current browser session's filter state (time block, source series, threshold, and data-quality filter). It is not sent to a server. A user can select a saved state to restore the controls that produced a comparison. |
| - **Finding boundary** appears beside the selected record and joins the record, selected threshold, data scope, and prohibited inference. This is a student-authored extension inspired by the interaction-provenance concepts in Narechania et al. (IEEE VIS 2024); it is not a claim that the application uses the ProvenanceWidgets code library. |
|
|
| ## Transformation record |
|
|
| 1. Read six source series and eight consecutive timestamps from `Traffic.csv`; preserve the source series identifier and timestamp. |
| 2. Convert occupancy rates from 0-1 to percentages and retain values to two decimal places. All included records are observed; no missing values are recoded as zero. |
| 3. Filter selected records by interval, source series, search term, data status, and threshold. |
| 4. Calculate a source-series baseline from displayed values. Flag an observed value when it is at or above the selected threshold. |
| 5. Render a heatmap and an accessible selected-record detail view. Explain the comparison; do not assign a cause. |
| 6. Keep at most five local-session control states. Do not collect, persist, or transmit user identifiers or interaction history. |
|
|
| ## Verification still required |
|
|
| - Test in a signed-out browser after deployment. |
| - Capture filter, selection, source-error, keyboard, colour-contrast, and mobile-layout evidence for Appendix A. |
| - Validate the project purpose with prospective users or a responsible proxy. |
|
|
| ## Sources |
|
|
| - [Hugging Face Static HTML Spaces](https://huggingface.co/docs/hub/spaces-sdks-static) |
| - [Hugging Face Dataset Cards](https://huggingface.co/docs/hub/datasets-cards) |
| - [Duyu Traffic benchmark collection](https://huggingface.co/datasets/Duyu/Time-Series-Forecasting-Benchmark-Datasets) |
|
|
