data trove bond girls

Overview

Source: /home/coolhand/html/datavis/data_trove/entertainment/film/bond_girls.csv

Saturn profiled 71 rows across 11 columns. The stats below are deterministic and machine-readable; the prose is a language-model interpretation of those stats (opt-in, added after the fact, never sees raw rows).

bond_girl_name categorical label

This column contains the names of Bond girls (female characters/actresses from the James Bond film franchise), serving as a near-unique label per row across 71 records. Cardinality is 70 out of 71 rows, with Léa Seydoux being the only duplicate (appearing twice, likely reflecting her appearances in both Spectre and No Time to Die). Entropy ratio of 0.999 confirms the distribution is almost perfectly uniform — every other name appears exactly once, triggering the long-tail alert despite the dataset being small.

Treatment: Use as a display label or join key to a Bond film lookup table; not suitable as a categorical feature due to near-unique cardinality.

actress_age numeric feature

This column records the age of an actress at the time of some event (likely a film role or award), spanning 21 to 51 years across 71 records. With only 20 unique values despite 71 rows, ages are heavily repeated, suggesting a small cast or repeated appearances by the same actresses. The distribution is moderately right-skewed (skew 0.97) with a single outlier at 51, while the bulk of values cluster between 24 and 33 (IQR = 9.0) around a mean of ~28.9 — consistent with known industry bias toward younger actresses.

Treatment: Use as-is or bin into age brackets; consider log-transform if used in regression given positive skew.

film_title categorical label

This column contains film titles from the James Bond franchise, covering at least 25 distinct films across 71 rows — indicating each film appears multiple times (averaging ~2.8 rows per title). The distribution is remarkably flat: the top value 'Thunderball' appears only 5 times (7.04% top_rate) and entropy_ratio is 0.984, meaning titles are spread almost uniformly across the dataset, suggesting rows likely represent individual scenes, characters, songs, or some other per-film sub-records rather than one row per film.

Treatment: Use as a grouping key for aggregations; do not encode as a high-cardinality feature without grouping logic.

film_release_year numeric feature

This column records the theatrical release year of films, spanning 1962 to 2021 across 71 rows. With only 25 unique values across 71 records, many films share a release year, suggesting clustering around certain production eras. The distribution is mildly right-skewed (skew 0.60) with a median of 1979 and mean of 1983, indicating the dataset leans toward older films, though at least one entry extends to 2021. The IQR of 30 years (1967–1997) confirms the bulk of titles fall in a mid-20th-to-late-20th-century range.

Treatment: Use as a numeric feature or bin into decade-level categories for modelling; consider interaction with other film metadata.

bond_actor categorical label

This column identifies which actor portrayed James Bond in each record, covering all 6 canonical Bond actors across 71 rows. Sean Connery leads with 23 appearances (32.4% of records), followed by Roger Moore (19) and Pierce Brosnan (12); George Lazenby and Timothy Dalton are notably underrepresented with only 3 records each, reflecting their shorter tenures. The entropy ratio of 0.88 indicates a moderately uneven but not extreme distribution. No nulls are present and cardinality is perfectly clean.

Treatment: One-hot encode or use as a grouping/stratification variable; note class imbalance for Lazenby and Dalton (3 records each) if modelling per-actor effects.

bond_actor_age numeric feature

This column records the age of the actor playing James Bond at the time of each film, spanning 71 observations (likely one per film or scene) across only 21 unique values — consistent with a small roster of actors each appearing in multiple entries. The distribution is remarkably symmetric and platykurtic (kurtosis −1.02), ranging from 30 to 58 with a mean and median both near 43, suggesting Bond actors tend to be cast and retained through their late-30s to late-40s with no strong skew toward younger or older talent.

Treatment: Use as-is for modelling; low cardinality (21 unique values) means binning or treating as ordinal may also be appropriate.

director_name categorical label

This column contains the names of film directors, almost certainly for the James Bond franchise — the 12 unique directors across 71 records are an exact match for the known Bond film series. Guy Hamilton leads with 14 films, followed by John Glen (11), Terence Young (10), and Lewis Gilbert (10), accounting for the bulk of the series. The entropy ratio of 0.917 is surprisingly high for only 12 categories, indicating the distribution is relatively flat rather than dominated by one director. No nulls are present.

Treatment: Use as a categorical grouping variable; with only 12 distinct values and no nulls, one-hot or ordinal encoding is straightforward.

box_office_actual_$ numeric numeric_target

This column records actual box office revenue (likely in millions of dollars) for 71 films, with only 25 unique values suggesting heavy rounding or binning of financial figures. The distribution is strongly right-skewed (skew = 1.92, kurtosis = 3.55): the median is $156.2M while the mean is pulled to $265.4M, and the max reaches $1,108.6M — nearly 5× the mean — with 4 flagged outliers (5.6% of rows) driving this spread. The IQR of $231.55M against a std of $235.3M confirms that a small number of blockbusters are distorting the central tendency.

Treatment: Log-transform before regression to reduce skew; investigate the 4 outliers (≥ upper fence) and confirm whether the 25-unique-value clustering reflects intentional rounding.

box_office_adjusted_2005_$ numeric feature

This column represents inflation-adjusted box office revenue (in millions of 2005 USD) for a small set of films or releases (n=71). Striking is the very low cardinality — only 24 unique values across 71 rows — suggesting heavy value repetition or bucketing, which is unusual for a continuous financial metric. The distribution is moderately right-skewed (skew=0.81) with a high outlier rate of 26.8% (19 of 71 observations), and values range from $250.9M to $943.5M, indicating this dataset likely covers only commercially successful titles.

Treatment: Investigate low cardinality (24 unique values in 71 rows) before use; if intentional bucketing, treat as ordinal; otherwise log-transform to reduce right skew before regression.

budget_actual_$ categorical feature

This column represents actual budget figures in dollars (likely millions), stored as strings rather than a numeric type — a likely ingestion or schema issue. With 23 unique values across 71 rows and an entropy ratio of 0.971, the distribution is nearly uniform, meaning no single budget tier dominates heavily (top value '7.0' appears only 8 times, ~11.3% of rows). Values range from small figures like '2.0' up to '34.0', suggesting a wide spread of project or department budgets. The categorical typing of a clearly numeric column is the primary surprise and will break any quantitative analysis.

Treatment: Cast to float, then assess distribution shape before deciding on scaling or log-transform for modelling.

budget_adjusted_2005_$ categorical feature

This column represents inflation-adjusted budget figures (in millions of 2005 USD), but it has been stored as a categorical type rather than numeric — a likely ingestion or typing error. With 71 rows and 25 unique values, cardinality is moderate, but an entropy ratio of 0.984 indicates near-uniform distribution with no dominant value (top value '41.9' appears only 5 times, a 7% top_rate). The decimal-formatted string values (e.g., '41.9', '91.5') confirm these are numeric quantities masquerading as categories.

Treatment: Cast to float64, verify unit scale (likely millions USD), then use directly or log-transform before modelling.