Pandas

Spike Detection in Time Series

Detect anomalies by comparing deltas to the median. Works for traffic, errors, costs - any metric with periodic measurements.

# Calculate period-over-period deltas
df['delta'] = df['value'].diff()

# Flag spikes: delta significantly above baseline
median_delta = df['delta'].median()
df['is_spike'] = df['delta'] > median_delta * 2

The median is robust to outliers. Multiplier of 2 catches significant deviations without false positives.

References

• pandas Documentation