notebooks/ToL_predicted-catalog_EDA.py

# ---
# jupyter:
#   jupytext:
#     formats: ipynb,py:percent
#     text_representation:
#       extension: .py
#       format_name: percent
#       format_version: '1.3'
#       jupytext_version: 1.15.2
#   kernelspec:
#     display_name: Python 3 (ipykernel)
#     language: python
#     name: python3
# ---

# %%
import pandas as pd
import seaborn as sns

sns.set_style("whitegrid")
sns.set(rc = {'figure.figsize': (10,10)})

# %%
df = pd.read_csv("../data/predicted-catalog.csv")

# %%
df.head()

# %%
df.info(show_counts = True)

# %% [markdown]
# The `train_small` is duplicates of `train`, so we will drop those to analyze the full training set plus val.

# %%
df = df.loc[df.split != 'train_small']

# %%
df.info(show_counts = True)

# %% [markdown]
# `predicted-catalog` doesn't have `train_small`, hence, it's a smaller file.

# %% [markdown]
# Original version had 10,436,521 entries; we expected loss of about 84K from the genera with label "unknown".
# We have still lost about 300K images from the original 10,092,530, but the pre-webdataset generation gained about 30K back from v3.2 (adding subspecies back in?).
#
# .......
#
# after webdataset generation:
#
# , but there's still another ~300K missing for this 10,065,576. It seems subspecies were not integrated back in under their species for this version, as we now have 269 less images than last time.
#
# Coverage for species and genus has also dropped by 269 and similar, resp.

# %% [markdown]
# ### Focus here on the difference between `predicted-catalog` and `catalog`, which is only in EOL data.

# %%
df.nunique()

# %% [markdown]
# There are 504,018 unique EOL page IDs (total 6,277,374 entries), compared to the 503,589 in the webdataset (total 6,250,420 images).

# %%
# Number that get dropped in webdataset
print(f"There are {6277374 - 6250420} less entries in the webdataset.")

# %% [markdown]
# Notice that we have 12 unique kingdoms, which we're sticking with.

# %%
df['kingdom'].value_counts()

# %% [markdown]
# There is 1 more member of `Plantae` predicted, but 27K more `Animalia`.

# %%
taxa = list(df.columns[9:16])
taxa

# %% [markdown]
# Check the number of images with all 7 taxonomic labels.

# %%
df_all_taxa = df.dropna(subset = taxa)
df_all_taxa[taxa].info(show_counts = True)

# %% [markdown]
# We have 8,482,197 entries with full taxonomic labels, compared to  8,455,243 in the webdataset, so 
# ### _**all**_ of our lost entries have all taxonomic ranks filled.

# %% [markdown]
# Let's add a column indicating the original data source so we can also get some stats by datasource, specifically focusing on EOL now.

# %%
# Add data_source column for easier slicing
df.loc[df['inat21_filename'].notna(), 'data_source'] = 'iNat21'
df.loc[df['bioscan_filename'].notna(), 'data_source'] = 'BIOSCAN'
df.loc[df['eol_content_id'].notna(), 'data_source'] = 'EOL'

# %% [markdown]
# First, check their unique class values (`common`).

# %%
df.loc[df['data_source'] == 'EOL', 'common'].nunique()

# %% [markdown]
# EOL number of unique classes is 439,910 in the webdataset, so we do lose 386...

# %% [markdown]
# Make `df_taxa` with just taxa columns (+ `common` & `data_source`) so it's smaller to process faster.

# %%
taxa_com = list(df.columns[9:17]) # taxa + common
taxa_com.insert(0, 'data_source')
df_taxa = df[taxa_com]

# %%
df_taxa.head()

# %% [markdown]
# Let's look a little closer at EOL.

# %%
inat21_df = df_taxa.loc[df_taxa.data_source == 'iNat21']
bioscan_df = df_taxa.loc[df_taxa.data_source == 'BIOSCAN']
eol_df = df_taxa.loc[df_taxa.data_source == 'EOL']

# %% [markdown]
# ### EOL

# %%
eol_df.info(show_counts = True)

# %%
eol_df.nunique()

# %% [markdown]
# It seems the webdataset loses 5 families, 39 genera, and 84 species. As noted above, 386 common labels are lost.

# %% [markdown]
# There are no missing species that get lost, as observed above, all lost entries have all taxonomic labels.

# %%
#number of unique 7-tuples in EOL
eol_df['duplicate'] = eol_df.duplicated(subset = taxa, keep = 'first')
eol_df_unique_taxa = eol_df.loc[~eol_df['duplicate']]

# %%
eol_df_unique_taxa.info(show_counts = True)

# %% [markdown]
# 400 unique taxa are lost.

# %% [markdown]
# Let's check the `species` length in EOL as well, we know there are some that have genus-species. And others with hybrids that get VERY long.
#
# Wonder if this is where we lose some?
# Quick check in the catalog notebook shows there are still plenty with full taxa there.

# %%