ML Mean, Median, and Mode

ML Mean, Median, and Mode: Measuring the Center of Data

Mean, median, and mode are the three core measures of central tendency in Machine Learning. Exploring them before training helps you spot skewed columns, outliers, and missing-value imputation strategies right away.

The mean is the arithmetic average, the median is the middle value after sorting, and the mode is the value that appears most often.

Computing Mean, Median, and Mode in Python

Python's built-in statistics module computes all three directly on small lists without any third-party install (numpy and scipy.stats scale the same logic to large arrays).

import statistics

salaries = [48000, 52000, 52000, 55000, 60000, 290000]

avg_salary = statistics.mean(salaries)
mid_salary = statistics.median(salaries)
common_salary = statistics.mode(salaries)

print(f"Mean:   ${avg_salary:,.0f}")    # $92,833 (skewed by 290k!)
print(f"Median: ${mid_salary:,.0f}")    # $53,500
print(f"Mode:   ${common_salary:,.0f}") # $52,000

Notice how a single 290000 executive salary pulls the mean to $92,833, even though five of six people earn $60,000 or less.

Choosing the Right Metric for Missing Values

When filling missing numeric values in an ML dataset, use the median for skewed features (like income or home prices), the mean for symmetric bell-shaped features, and the mode for categorical labels.

Common Mistakes to Avoid

• Using the mean on skewed data: Averages mislead stakeholders on income, latency, and price distributions with heavy right tails.
• Forgetting multimode() when ties exist: statistics.mode() returns the first tie; statistics.multimode() returns all tied top values.
• Imputing missing values before train/test split: Computing dataset medians across the entire table leaks test-set distribution into training.

Frequently Asked Questions

What happens to the median when a list has an even number of items?

Python sorts the list and returns the average of the two middle numbers (for [52000, 55000], the median is 53500.0).

Can mode work on non-numeric strings?

Yes. statistics.mode(["red", "blue", "red"]) returns "red", making mode ideal for categorical features.

How do NumPy and SciPy compute these for large arrays?

Use np.mean(arr) and np.median(arr) in NumPy, and scipy.stats.mode(arr) for mode on NumPy arrays.