Sliding Window Anomaly Detection | Dataford Interview Questions

Implement an algorithmic solution to identify anomalies within a sliding time-window stream.

Given time-ordered events (timestamp, value), return the zero-based indices whose values are anomalous compared with prior events inside the time window. Use the population standard deviation and classify an event when its absolute z-score is greater than threshold after at least min_history prior events exist. If the standard deviation is zero, any different value is anomalous. Implement detect_anomalies(events, window, threshold, min_history) and return a list of indices.

Constraints

1 <= len(events) <= 5000
Events are sorted by nondecreasing timestamp
Each event is a two-element pair [timestamp, value]
0 <= timestamp <= 10^12
-10^9 <= value <= 10^9
0 < window <= 10^12
0 <= threshold
1 <= min_history <= len(events)

Function Signature

def detect_anomalies(events, window, threshold, min_history):

Your question is Sliding Window 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 an algorithmic solution to identify anomalies within a sliding time-window stream.

Constraints

1 <= len(events) <= 5000
Events are sorted by nondecreasing timestamp
Each event is a two-element pair [timestamp, value]
0 <= timestamp <= 10^12
-10^9 <= value <= 10^9
0 < window <= 10^12
0 <= threshold
1 <= min_history <= len(events)

Function Signature

def detect_anomalies(events, window, threshold, min_history):

Your question is Sliding Window 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.