Benj-samurai commited on
Commit
e065c18
·
verified ·
1 Parent(s): 9049dfe

Update README.md

Browse files

docs: add complete dataset card with YAML metadata, columns, usage & license

Files changed (1) hide show
  1. README.md +119 -53
README.md CHANGED
@@ -1,19 +1,33 @@
1
- # Strava Master & Weekly Summary Dataset
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2
 
3
- Personal Strava export processed on **2025-05-04**
4
- Provides per-activity master table and weekly aggregations with intensity metrics.
5
 
6
  ---
7
 
8
  ## Files
9
 
10
  | File | Rows | Description |
11
- |------|------|-------------|
12
- | `strava_master_enhanced.parquet` | ~N | 1 row = 1 activity, cleaned & enriched |
13
- | `weekly_sport.parquet` | ~N | Weekly totals by sport (Run/Ride/Swim …) |
14
- | `weekly_category.parquet` | ~N | Weekly totals by intensity zone |
15
 
16
- *(rows will auto-update, no needに手入力可か削除でもOK)*
17
 
18
  ---
19
 
@@ -21,71 +35,123 @@ Provides per-activity master table and weekly aggregations with intensity metric
21
 
22
  | Column | `dtype` | Description |
23
  |--------|---------|-------------|
24
- | `activity_id` | `int64` | Unique Strava activity ID |
25
- | `name` | `string` | Activity title on Strava |
26
- | `sport` | `category` | Sport type (Run / Ride / Swim / Walk / ) |
27
- | `date` | `datetime64[ns]` | Local start time |
28
- | `distance_km` | `float32` | Distance (metres**km**) |
29
- | `elapsed_hr` | `float32` | Elapsed time (sec → **h**; incl. stops) |
30
- | `moving_hr` | `float32` | Moving time (sec **h**) |
31
- | `elevation_gain_m` | `float32` | Positive elevation gain (m) |
32
- | `elevation_loss_m` | `float32` | Negative elevation (descent, m) |
33
- | `average_speed_kph` | `float32` | Moving speed (km h) |
34
- | `max_speed_kph` | `float32` | Max speed (km h) |
35
- | `average_hr` | `float32` | Avg heart-rate (bpm); *NaN* if no sensor |
36
  | `max_hr` | `int16` | Max heart-rate (bpm) |
37
- | `average_cadence` | `float32` | Avg cadence (rpm); bike-runs mixed |
38
  | `max_cadence` | `int16` | Max cadence (rpm) |
39
- | `intensity_level` | `float32` | Avg HR ÷ **LTHR (165 bpm)** 0.50–1.10 |
40
- | `training_category` | `category` | HR zone label: `Z1-2`, `Z3`, `Z4`, `Z5`, `NoHR` |
41
- | `trimp` | `float32` | Banister TRIMP = `moving_hr × intensity_level × 50` |
42
- | `pace_min_per_km` | `float32` | Pace (min km-¹); *NaN* for non-run sports |
43
- | `commute` | `boolean` | Marked as commute on Strava |
44
- | `gear` | `string` | Bike / Shoes used (if set) |
45
  | `calories_kcal` | `float32` | Calories reported by Strava |
 
 
 
 
 
 
46
  | `weather` | `string` | Weather summary (if available) |
47
- | `temperature_c` | `float32` | Avg temperature (°C) |
48
- | `relative_effort` | `Int32` | Strava Relative Effort score |
49
- | `filename` | `string` | Original FIT/GPX filename (metadata only) |
50
- | `gpx_path` | `string | None` | Path to GPX in `routes/` (not included if `None`) |
51
-
52
- > **Note**
53
- > * All numeric time/ distance columns are converted to SI units (hours & kilometres) for easy aggregation.
54
- > * Empty sensor data (e.g. HR-less activities) are stored as **`NaN`** so they don’t skew means.
55
- > * `training_category` bins: Z1-2 < 0.79, Z3 < 0.89, Z4 < 0.95, Z5 ≥ 0.95 of LTHR.
56
- > * `gpx_path` points to a GeoJSON-converted route if the original GPX was processed; raw GPS files are **not** in the repo for privacy.
 
 
 
 
 
 
 
 
 
57
 
58
 
59
  ---
60
 
61
  ## Processing pipeline
62
 
63
- 1. **Header translation** : JP CSV → EN headers (custom script)
64
- 2. **Unit conversion** : distance mkm, time s→h
65
- 3. **Intensity & TRIMP** : LTHR = 165 bpm, zones Z1-2/3/4/5
66
- 4. **Weekly aggregation** via pandas `groupby`
 
 
67
 
68
- Full code is available in `notebooks/` and `make_dataset.py` inside the GitHub repo.
 
69
 
70
  ---
71
 
72
- ## Usage example
73
 
74
  ```python
75
  from datasets import load_dataset
76
- ds = load_dataset("Benj-samurai/strava_dataset",
77
- data_files="strava_master_enhanced.parquet")
 
 
 
 
78
  df = ds["train"].to_pandas()
79
- print(df.head())
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
80
  ```
81
 
82
- ## License
83
 
84
- Data © <Yuki Asai> 2025 — released under CC BY-NC 4.0.
85
- Commercial use forbidden. Remove GPS data before public redistribution.
 
86
 
87
- ## Privacy notes
88
 
89
- - **Raw FIT/GPX files are _not_ included.**
90
- - **No exact home coordinates**
91
- All activity start points are shifted by 200 m to obscure the true home location.
 
 
 
 
 
 
 
1
+ ---
2
+ dataset_name: strava_master_dataset
3
+ pretty_name: Strava Master Dataset
4
+ license: cc-by-nc-4.0
5
+ task_categories:
6
+ - time-series-forecasting # ← ここを修正
7
+ tags:
8
+ - running
9
+ - cycling
10
+ - wearable
11
+ language:
12
+ - en
13
+ ---
14
+
15
+ # Strava Master Dataset ― Benj-samurai
16
 
17
+ > **Personal multi-sport training log** exported from Strava (JP CSV) and processed into a clean, analysis-ready Parquet table + weekly summaries.
18
+ > Covers **〔開始日 終了日〕**, total **〔活動件数〕 activities** across **〔種目数〕 sports**.
19
 
20
  ---
21
 
22
  ## Files
23
 
24
  | File | Rows | Description |
25
+ |------|-----:|-------------|
26
+ | `my_strava_dataset/strava_master_enhanced.parquet` | 〔n〕 | Master table — 1 row = 1 activity |
27
+ | `my_strava_dataset/weekly_sport.parquet` | 〔m〕 | Weekly totals by *year–week × sport* |
28
+ | `my_strava_dataset/weekly_category.parquet` | 〔k〕 | Weekly totals by intensity zone |
29
 
30
+ *(Parquet compact, schema-aware, loadable via `datasets.load_dataset`)*
31
 
32
  ---
33
 
 
35
 
36
  | Column | `dtype` | Description |
37
  |--------|---------|-------------|
38
+ | `activity_id` | `int64` | Strava Activity ID |
39
+ | `name` | `string` | Activity title as saved on Strava |
40
+ | `sport` | `category` | Sport type (`Run`, `Ride`, `Swim`, `Walk`, …) |
41
+ | `date` | `datetime64[ns]` | Local activity start time |
42
+ | `distance_km` | `float32` | Distance in **kilometres** (raw m → km) |
43
+ | `elapsed_hr` | `float32` | Elapsed time incl. pauses **hours** (raw sec → h) |
44
+ | `moving_hr` | `float32` | Moving time (in-motion) **hours** |
45
+ | `elevation_gain_m` | `float32` | Total positive elevation gain (m) |
46
+ | `elevation_loss_m` | `float32` | Total negative elevation (m) |
47
+ | `average_speed_kph` | `float32` | Moving speed (km h⁻¹) |
48
+ | `max_speed_kph` | `float32` | Max speed (km h⁻¹) |
49
+ | `average_hr` | `float32` | Avg heart-rate (bpm) *NaN if no sensor* |
50
  | `max_hr` | `int16` | Max heart-rate (bpm) |
51
+ | `average_cadence` | `float32` | Avg cadence (rpm) |
52
  | `max_cadence` | `int16` | Max cadence (rpm) |
53
+ | `average_power` | `float32` | Avg power (W); bike only |
54
+ | `max_power` | `int16` | Peak power (W) |
 
 
 
 
55
  | `calories_kcal` | `float32` | Calories reported by Strava |
56
+ | `training_category` | `category` | HR zone label `Z1-2 / Z3 / Z4 / Z5 / NoHR` |
57
+ | `intensity_level` | `float32` | Avg HR ÷ LTHR (165 bpm) |
58
+ | `trimp` | `float32` | Banister TRIMP (`moving_hr × intensity_level × 50`) |
59
+ | `commute` | `boolean` | Marked as commute on Strava |
60
+ | `filename` | `string` | Original FIT/GPX filename (meta only) |
61
+ | `gear` | `string` | Bike / shoes used (if set) |
62
  | `weather` | `string` | Weather summary (if available) |
63
+ | `temperature_c` | `float32` | Avg temp (°C) |
64
+ | `flagged` | `boolean` | Strava flagged activity |
65
+ | `year` | `int16` | Calendar year (`date`). fast grouping |
66
+ | `month` | `int16` | Calendar month (1–12) |
67
+ | `week` | `int16` | ISO week number (1–53) |
68
+ | `year_month` | `string` | `"YYYY-MM"` label for plotting |
69
+ | `week_start` | `datetime64[ns]` | Monday of ISO week (analysis helper) |
70
+ | `sport_weekly_id` | `string` | Composite key `year_week-sport` |
71
+ | `distance_ratio` | `float32` | Share of weekly distance (per sport) |
72
+ | `pace_min_per_km` | `float32` | Pace (min km⁻¹); NaN for non-run |
73
+ | `grade_adjusted_pace` | `float32` | GAP (min km⁻¹); run only |
74
+ | `dirt_distance_km` | `float32` | Unpaved distance (km) |
75
+ | `total_cycles` | `int32` | Swim strokes / pedal revs where available |
76
+ | `route_hash` | `string` | MD5 of polyline (GPS privacy) |
77
+ | `gpx_path` | `string \| None` | Optional GeoJSON path file |
78
+
79
+ > *All numeric distance/time columns are converted to km / hours and stored in
80
+ > low-memory float32/int16 where possible.
81
+ > Empty sensor data are kept as **`NaN`** so they don’t skew means.*
82
 
83
 
84
  ---
85
 
86
  ## Processing pipeline
87
 
88
+ 1. **Export**: Strava JP CSV (`activities.csv`)
89
+ 2. **Header translation** JPEN (`translate_headers.py`)
90
+ 3. Unit conversion (m→km, s→h) & dtype down-cast (`clean_master.ipynb`)
91
+ 4. **Intensity & TRIMP**: LTHR = 165 bpm, Banister formula
92
+ 5. Weekly aggregations (`weekly_summary.ipynb`)
93
+ 6. Saved as Parquet, tracked via **Git LFS** (compact & diff-friendly)
94
 
95
+ Code & notebooks live in the companion GitHub repo:
96
+ <https://github.com/Benj-samurai/Strava-Dataset-PRJ>
97
 
98
  ---
99
 
100
+ ## Usage
101
 
102
  ```python
103
  from datasets import load_dataset
104
+
105
+ ds = load_dataset(
106
+ "Benj-samurai/strava_dataset",
107
+ data_files="my_strava_dataset/strava_master_enhanced.parquet",
108
+ streaming=False, # True = stream without download
109
+ )
110
  df = ds["train"].to_pandas()
111
+
112
+ # quick EDA
113
+ weekly_km = (
114
+ df.assign(year_week=df["date"].dt.to_period("W"))
115
+ .groupby(["year_week", "sport"])["distance_km"].sum()
116
+ )
117
+ print(weekly_km.tail())
118
+ ```
119
+
120
+ ## Privacy & Personal-use License
121
+
122
+ - **Raw FIT/GPX files are _not_ included.**
123
+ - **Activity start coordinates are jittered ≥ 200 m** to obscure the true home location.
124
+ - Released under **CC BY-NC 4.0** – non-commercial use, attribution required.
125
+ If you wish to use the data commercially, please contact the author first.
126
+
127
+ ---
128
+
129
+ ## Citation
130
+
131
+ ```bibtex
132
+ @misc{asai2025strava,
133
+ author = {Asai, Benj-samurai},
134
+ title = {Strava Master Dataset},
135
+ year = {2025},
136
+ howpublished = {\url{https://huggingface.co/datasets/Benj-samurai/strava_dataset}},
137
+ note = {Version {{\today}}}
138
+ }
139
  ```
140
 
141
+ ## Changelog
142
 
143
+ | Date | Version | Notes |
144
+ |------------|---------|--------------------------|
145
+ | 2025-05-06 | v1.0 | Initial public release |
146
 
147
+ ---
148
 
149
+ ### 使い方
150
+
151
+ 1. HF Hub ページの **Dataset card** タブ **Edit** を開く
152
+ 2. 上の Markdown を貼り付ける
153
+ 3. 〔 〕部分を実際の値に置換して **Commit**
154
+ *行数* は手元で `len(df)`、週レコードは `len(weekly_sport)` などで確認できます。
155
+
156
+ これで “列定義・処理手順・使用例・ライセンス” を網羅したリッチな Dataset Card になります。
157
+ 追記やレイアウト調整はお好みでどうぞ!