arrow_prepare_dataset.py

## Read Palmer Station Penguin dataset from GitHub
import pandas as pd
df = pd.read_csv("https://raw.githubusercontent.com/allisonhorst/"
                 "palmerpenguins/47a3476d2147080e7ceccef4cf70105c808f2cbf/"
                 "data-raw/penguins_raw.csv")
                 # Increase dataset to 1m rows and reset index
df = df.sample(1_000_000, replace=True).reset_index(drop=True)


# Update sample number (0 to 999'999)
df["Sample Number"] = df.index
# Add some random variation to numeric columns
df[["Culmen Length (mm)", "Culmen Depth (mm)", 
    "Flipper Length (mm)", "Body Mass (g)"]] = df[["Culmen Length (mm)", "Culmen Depth (mm)", 
                                                   "Flipper Length (mm)", "Body Mass (g)"]] \
                                               + np.random.rand(df.shape[0], 4)

# Create dataframe where missing numeric values are filled with zero
df_nonan = df.copy()
df_nonan[["Culmen Length (mm)", "Culmen Depth (mm)", 
          "Flipper Length (mm)", "Body Mass (g)"]] = df[["Culmen Length (mm)", "Culmen Depth (mm)", 
                                                         "Flipper Length (mm)", "Body Mass (g)"]].fillna(0)