This lesson's interactive features are locked, please to keep using them

Type Normalization

Raw data is so often the wrong type. Numbers, dates, and booleans often arrive as strings: "100", "2024-01-15", "Yes".

Type normalization means converting values into the types that make the most sense. Say you want the average shipping cost from an order export, but the values are strings:

df["shipping_cost"] = ["7.25", "12.00", "4.50"]
df["shipping_cost"].mean()
# TypeError

Converting to Numeric

pd.to_numeric() converts strings to numbers:

# Convert a column to numeric
df["shipping_cost"] = pd.to_numeric(df["shipping_cost"], errors="coerce")

# Now you can do math
df["shipping_cost"].mean()

The errors="coerce" option turns unconvertible values into NaN instead of crashing your code.

Checking Types

Before you convert anything, check what you actually have:

# Show types of all columns
print(df.dtypes)

# Check a specific column
print(df["shipping_cost"].dtype)

Assignment

Complete the split_healthy_unhealthy_devices function. It accepts a DataFrame and a threshold value, then converts the values in the temperature column to numbers and returns two DataFrames: healthy devices below the threshold and unhealthy devices at or above the threshold.

Create a copy of the DataFrame to avoid modifying the original.
Convert the temperature column to numeric and coerce the errors.
Filter devices where temperature is below the threshold into a "healthy" DataFrame.
Filter devices where temperature is at or above the threshold into an "unhealthy" DataFrame.
Return both DataFrames as a tuple: healthy, then unhealthy.