Anomaly Detection with Rolling Windows | Dataford Interview Questions

Your question is Anomaly Detection with Rolling Windows. Start with the requirements on the right.

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

Problem

Given a stream of financial transactions, write an efficient algorithm to detect anomalies based on rolling statistical windows.

Implement detect_anomalies(transactions, window_size, z_threshold). For each transaction after the first complete preceding window, calculate the window's mean and sample standard deviation. Mark the transaction as anomalous when its absolute z-score is strictly greater than z_threshold. Return anomalous zero-based indices. If the window standard deviation is zero, mark the value anomalous only when it differs from the window mean. Do not evaluate transactions before a complete window exists.

Input values are numeric, window_size is an integer of at least 2, and z_threshold is nonnegative.

Constraints

2 <= window_size <= len(transactions) or len(transactions) may be smaller than window_size
0 <= len(transactions) <= 100000
Each transaction is an integer or floating-point number
0 <= z_threshold
Use the sample standard deviation for each preceding window

Function Signature

def detect_anomalies(transactions, window_size, z_threshold):

Your question is Anomaly Detection with Rolling Windows. Start with the requirements on the right.

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

Given a stream of financial transactions, write an efficient algorithm to detect anomalies based on rolling statistical windows.

Input values are numeric, window_size is an integer of at least 2, and z_threshold is nonnegative.

Constraints

2 <= window_size <= len(transactions) or len(transactions) may be smaller than window_size
0 <= len(transactions) <= 100000
Each transaction is an integer or floating-point number
0 <= z_threshold
Use the sample standard deviation for each preceding window

Function Signature

def detect_anomalies(transactions, window_size, z_threshold):