Streaming Anomaly Detection | Dataford Interview Questions

Implement a function to process a stream of data and detect anomalies.

The function receives numeric observations, a rolling window size, and a z-score threshold. For each observation after the window is full, compare it with the preceding window and return the indices whose absolute z-score exceeds the threshold. If the preceding window has zero variance, any different value is anomalous; the first window_size observations cannot be flagged.

Input/output: detect_anomalies(stream, window_size, threshold) receives a list of numbers, a positive integer, and a positive number. It returns a list of anomalous zero-based indices.

Constraints

1 <= len(stream) <= 100000
1 <= window_size <= len(stream)
threshold > 0
-10^9 <= stream[i] <= 10^9
Use the population standard deviation of the preceding window
Anomaly detection uses a strict comparison: z-score > threshold

Function Signature

def detect_anomalies(stream, window_size, threshold):

Your question is Streaming Anomaly Detection. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem

Implement a function to process a stream of data and detect anomalies.

Input/output: detect_anomalies(stream, window_size, threshold) receives a list of numbers, a positive integer, and a positive number. It returns a list of anomalous zero-based indices.

Constraints

1 <= len(stream) <= 100000
1 <= window_size <= len(stream)
threshold > 0
-10^9 <= stream[i] <= 10^9
Use the population standard deviation of the preceding window
Anomaly detection uses a strict comparison: z-score > threshold

Function Signature

def detect_anomalies(stream, window_size, threshold):

Your question is Streaming Anomaly Detection. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.