quirky geothermal

Short free-text names for places, almost certainly hot springs and geysers given that 'hot_spring' (3393) and 'geyser' (1083) dominate top_words. The column is highly diverse (8316 unique of 8776) but multilingual — top values mix Arabic, English, Spanish, and Turkish — and 25.3% of entries are all-caps, which would break naive string matching. There are 460 duplicates (5.2%) including 40 repeats of a single Arabic name, suggesting the same feature recorded multiple times. Treatment: Normalize case and unicode, then keep as a descriptive label; do not use as a join key given duplicates and language mix. high · anthropic:claude-opus-4-7

n

8,776

nulls

0 (0.0%)

unique

8,316

len_min

1

len_max

60

len_mean

17.1

len_median

19

len_p95

24

word_mean

2.16

word_median

2

n_empty

0

n_duplicates

460

duplicate_rate

0.05242

vocab_size

9,610

readability_flesch_mean

101.6

emoji_rate

0

url_rate

0

one_word_rate

0.1508

allcaps_rate

0.2525

boilerplate_rate

0

This is a latitude coordinate column, with values spanning -54.68 to 70.96 and a median of 40.86 placing most records in the northern hemisphere. The distribution is heavily left-skewed (skew -2.36, kurtosis 6.28) with 889 outliers (10.1%), reflecting a long southern-hemisphere tail relative to a northern-clustered core. Near-unique values (8758/8776) indicate point-level geolocations rather than coarse bins. Treatment: Pair with longitude for geospatial features (e.g., bin, cluster, or compute distances) rather than treating as a standalone numeric. high · anthropic:claude-opus-4-7

n

8,776

nulls

0 (0.0%)

unique

8,758

min

-54.68

max

70.96

mean

34.77

median

40.86

std

18.12

q1

32.31

q3

44.53

iqr

12.22

skew

-2.36

kurtosis

6.277

n_outliers

889

outlier_rate

0.1013

zero_rate

0

This column is almost certainly geographic longitude in decimal degrees: values span -176.63 to 179.36, with mean -23.99 and median -21.23 sitting plausibly within the valid [-180, 180] range. The wide IQR of 154.96 and std of 89.39 indicate global coverage rather than a regional dataset, and near-uniqueness (8747 unique of 8776) suggests each row is a distinct location. No nulls, no zeros, and no flagged outliers. Treatment: Pair with latitude for geospatial features; avoid treating as a plain scalar in models due to wraparound at ±180. high · anthropic:claude-opus-4-7

n

8,776

nulls

0 (0.0%)

unique

8,747

min

-176.6

max

179.4

mean

-23.99

median

-21.23

std

89.39

q1

-110.8

q3

44.15

iqr

155

skew

0.4043

kurtosis

-1.179

n_outliers

0

outlier_rate

0

zero_rate

0

This is a country code field (ISO-2 style values like IQ, TW, MX, DE, RU, JP) that is effectively empty: 99.91% of the 8776 rows are null, leaving only 8 observed values across 6 distinct codes. The non-null distribution is too sparse to be meaningful, though IQ appears 3 times and accounts for 37.5% of present values. With this null rate, any apparent signal is noise. Treatment: Drop the column; null rate of 99.91% leaves nothing to model. high · anthropic:claude-opus-4-7

n

8,776

nulls

8,768 (99.9%)

unique

6

top_value

IQ

top_rate

0.375

cardinality

6

entropy

2.406

entropy_ratio

0.9306

Binary categorical column distinguishing two hydrothermal feature types: hot_spring (7082 rows, ~80.7%) and geyser (1694 rows). No nulls and only 2 unique values, so the field is clean but imbalanced roughly 4:1 toward hot_spring. Entropy ratio of 0.71 reflects that skew rather than any data quality issue. Treatment: One-hot or boolean-encode; stratify splits to preserve the ~4:1 class balance. high · anthropic:claude-opus-4-7

n

8,776

nulls

0 (0.0%)

unique

2

top_value

hot_spring

top_rate

0.807

cardinality

2

entropy

0.7078

entropy_ratio

0.7078

A free-text temperature field that is 98.31% null, with only 148 populated rows out of 8776. The dominant value is the descriptive string 'hot' (76 occurrences, 51.35% of populated rows), while the remaining entries are numeric strings like '90', '100', '21' — indicating a mix of qualitative and quantitative encodings with no consistent unit. Cardinality is 56 with entropy ratio 0.64, so the long tail is sparse but varied. Treatment: Drop or treat as missing-by-default; if retained, normalize units and split numeric vs categorical encodings before use. high · anthropic:claude-opus-4-7

n

8,776

nulls

8,628 (98.3%)

unique

56

top_value

hot

top_rate

0.5135

cardinality

56

entropy

3.734

entropy_ratio

0.643

This appears to be a Wikipedia article reference column, with values formatted as language-prefixed page titles (e.g., 'en:Olympic Hot Springs', 'ja:鉢形駅') spanning multiple languages including English, Japanese, Russian, and Icelandic. The column is 98.52% null with only 124 unique values across 8776 rows, and entropy ratio of 0.996 indicates the few populated entries are nearly all distinct. The top value appears just 3 times (0.023 rate), confirming no meaningful concentration. Treatment: Drop or retain only as a reference link; too sparse and high-cardinality for modelling. high · anthropic:claude-opus-4-7

n

8,776

nulls

8,646 (98.5%)

unique

124

top_value

en:Olympic Hot Springs

top_rate

0.02308

cardinality

124

entropy

6.924

entropy_ratio

0.9957

Free-text descriptions, likely for hot spring or geothermal site entries, present on only ~1.94% of the 8776 rows (null_rate 0.9806). Among the 170 non-null entries there are 141 unique values with entropy ratio 0.966, so almost every description is bespoke; the most common string ('Mud geyser created from recent seismic activity') still only repeats 12 times (top_rate 0.0706). Languages are mixed — English, Japanese (熱海七湯), Russian, French — which will complicate any text processing. Treatment: Treat as multilingual free text: language-detect then tokenize/embed, but expect ~98% missingness to limit usefulness as a feature. high · anthropic:claude-opus-4-7

n

8,776

nulls

8,606 (98.1%)

unique

141

top_value

Mud geyser created from recent seismic activity

top_rate

0.07059

cardinality

141

entropy

6.894

entropy_ratio

0.9656

Binary yes/no flag indicating whether something is intermittent, but it is essentially absent: 97.89% of the 8,776 rows are null, leaving only 185 populated values. Among those few, 'no' dominates at 91.9% (170 vs 15 'yes'), so the column carries almost no usable signal. Treatment: Drop or treat as a sparse indicator; null rate too high for direct modelling. high · anthropic:claude-opus-4-7

n

8,776

nulls

8,591 (97.9%)

unique

2

top_value

no

top_rate

0.9189

cardinality

2

entropy

0.406

entropy_ratio

0.406

This is a low-cardinality categorical access flag (5 distinct values: yes, customers, private, no, permissive) — likely an OSM-style access tag indicating who may use a feature. It is overwhelmingly null at 98.99%, leaving only 89 observed values across 8,776 rows. The non-null distribution is unusually flat (entropy ratio 0.945), with the modal value 'yes' accounting for just 26.97% of present values. Treatment: Treat missing as its own category ('unspecified') before encoding, given 99% nulls. high · anthropic:claude-opus-4-7

n

8,776

nulls

8,687 (99.0%)

unique

5

top_value

yes

top_rate

0.2697

cardinality

5

entropy

2.194

entropy_ratio

0.9451

This is an OpenStreetMap-style `tourism` tag classifying points of interest (attraction, hotel, camp_site, viewpoint, etc.). It is almost entirely empty — 97.46% null across 8776 rows — and among the 223 populated entries, `attraction` dominates at 87% (194 records), leaving the other 9 categories as long-tail singletons. A stray `yes` value (7 rows) suggests inconsistent tagging upstream. Treatment: Drop or collapse to a binary is_attraction flag; too sparse and skewed for direct use as a feature. high · anthropic:claude-opus-4-7

n

8,776

nulls

8,553 (97.5%)

unique

10

top_value

attraction

top_rate

0.87

cardinality

10

entropy

0.9127

entropy_ratio

0.2747

This is almost certainly the OpenStreetMap object identifier: every one of the 8776 rows is unique with no nulls or zeros, and values span 27750092 to 13535658843, the range typical of OSM IDs. The distribution is broad (IQR ~9.94e9) and slightly left-skewed (-0.30) with flat kurtosis (-1.43), consistent with IDs accumulated across OSM history rather than a meaningful numeric feature. No outliers were flagged, which is expected for an identifier. Treatment: Use as a join key to OSM; do not feed into models as a numeric feature. high · anthropic:claude-opus-4-7

n

8,776

nulls

0 (0.0%)

unique

8,776

min

2.775e+07

max

1.354e+10

mean

7.186e+09

median

8.263e+09

std

4.374e+09

q1

1.334e+09

q3

1.128e+10

iqr

9.942e+09

skew

-0.3011

kurtosis

-1.43

n_outliers

0

outlier_rate

0

zero_rate

0

This column records the OpenStreetMap geometry type, taking only two values across 8,776 rows: 'node' (6,705) and 'way' (2,071). 76.4% of records are nodes, giving an entropy ratio of 0.79 — moderately imbalanced but no nulls or rare categories. Treatment: One-hot or binary-encode before modelling. high · anthropic:claude-opus-4-7

n

8,776

nulls

0 (0.0%)

unique

2

top_value

node

top_rate

0.764

cardinality

2

entropy

0.7883

entropy_ratio

0.7883