---
pretty_name: YouTube Trending Video Statistics 2019-04-15 to 2020-04-15 TsFile
language:
- en
tags:
- tabular
- youtube
- tsd
- trending videos
- tsfile
- modality:timeseries
modality:
- timeseries
size_categories:
- 1M<n<10M
---

# youtube_trending_2019_2020

This repository contains a TsFile conversion of the Hugging Face dataset
[`jettisonthenet/timeseries_trending_youtube_videos_2019-04-15_to_2020-04-15`](https://huggingface.co/datasets/jettisonthenet/timeseries_trending_youtube_videos_2019-04-15_to_2020-04-15).

The source dataset card / Viewer identifies the dataset as a YouTube trending
video time-series dataset. The source repository contains `videostats.csv` and
the Hugging Face Viewer exposes the `default/train` split with about 1.54M rows.
The locally downloaded source file used for this conversion contains 1,541,128
rows.

## Original Dataset

- Original dataset: <https://huggingface.co/datasets/jettisonthenet/timeseries_trending_youtube_videos_2019-04-15_to_2020-04-15>
- Source file used for conversion: `videostats.csv`
- Split used: `train`
- Local downloaded rows: 1,541,128
- Distinct `ytvideoid` values in the downloaded file: 11,369
- Time range in the downloaded file: `2019-04-15 04:42:49` to `2020-04-14 23:31:06`

The original columns are:

| Column | Description used in this conversion |
|---|---|
| `videostatsid` | Source row/statistics identifier |
| `ytvideoid` | YouTube video identifier |
| `views` | View count |
| `comments` | Comment count |
| `likes` | Like count |
| `dislikes` | Dislike count |
| `timestamp` | Source timestamp |

The source dataset does not contain video files, frames, audio, thumbnails, or
other media binaries. `ytvideoid` is an identifier only.

## TsFile Conversion

The conversion was performed from the source CSV into staged Parquet and then
into TsFile using the repository workflow.

Generated files:

- `youtube_trending_2019_2020_1.tsfile`
- `youtube_trending_2019_2020_2.tsfile`

Schema-level conversion choices:

- `timestamp` was parsed into the TsFile `Time` column at millisecond precision.
- The original timestamp value was preserved as `source_timestamp_ms`.
- `ytvideoid` was declared as a TAG, so each YouTube video ID is represented as
  a separate device series.
- `views`, `comments`, `likes`, and `dislikes` were retained as `INT64` fields.
- `videostatsid` was retained as an `INT64` field for traceability.
- No original source columns were dropped.

Duplicate timestamp handling:

The source CSV contains 89 duplicate `(ytvideoid, timestamp)` pairs. TsFile
requires timestamps within one device series to be unique/ordered for reliable
import. For those duplicate pairs only, the converted `Time` value was
disambiguated by adding a zero-based millisecond offset ordered by
`videostatsid`. The unadjusted source timestamp remains available in
`source_timestamp_ms`.

## Validation

The staged Parquet file contains:

- 1,541,128 rows
- 11,369 distinct `ytvideoid` tag values
- 0 duplicate `(ytvideoid, Time)` pairs after conversion

Both generated `.tsfile` files were validated locally as present and non-empty.