metadata
tags:
- time series
- time series classification
- monster
- count
license: cc-by-4.0
pretty_name: Traffic
size_categories:
- 1M<n<10M
Part of MONSTER: https://arxiv.org/abs/2502.15122.
| Traffic | |
|---|---|
| Category | Count |
| Num. Examples | 1,460,968 |
| Num. Channels | 1 |
| Length | 24 |
| Sampling Freq. | hourly |
| Num. Classes | 7 |
| License | CC BY 4.0 |
| Citations | [1] |
Traffic consists of hourly traffic counts at various locations in the state of NSW, Australia [1]. The processed dataset contains 1,460,968 (univariate) time series, each of length 24 (i.e., representing 24 hours of data per time series). The data comes from automatic traffic counting sensors at different locations. The task is to predict the day of the week based on the time series of counts. The dataset has been split into stratified random cross-validation folds.
[1] Transport for NSW. (2023). NSW road traffic volume counts hourly. https://opendata.transport.nsw.gov.au/dataset/nsw-roads-traffic-volume-counts-api, CC BY 4.0.