Anomaly detection basics in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

What is the main goal of anomaly detection?

In simple terms, what does anomaly detection try to find in data?

A To find data points that are very different from the rest

B To find patterns that happen very often in the data

C To group data points into clusters based on similarity

D To predict future values based on past data

❓ Predict Output

intermediate

Output of simple anomaly score calculation

Given the following Python code that calculates anomaly scores as the absolute difference from the mean, what is the output list?

ML Python

data = [10, 12, 10, 13, 100]
mean = sum(data) / len(data)
anomaly_scores = [abs(x - mean) for x in data]
print([round(score, 2) for score in anomaly_scores])

A [18.0, 16.0, 18.0, 15.0, 72.0]

B [21.0, 19.0, 21.0, 18.0, 69.0]

C [19.0, 17.0, 19.0, 16.0, 71.0]

D [19.0, 17.0, 18.0, 16.0, 71.0]

❓ Model Choice

advanced

Best model choice for anomaly detection in high-dimensional data

You have a dataset with many features (dimensions) and want to detect anomalies. Which model is best suited?

A K-Nearest Neighbors (KNN) based anomaly detection

B Principal Component Analysis (PCA) based anomaly detection

C Linear Regression

D Decision Tree Classifier

❓ Metrics

advanced

Which metric is best to evaluate anomaly detection performance?

You have a model that flags anomalies. Which metric best measures how well it finds true anomalies without too many false alarms?

A Precision

B Accuracy

C Mean Squared Error

D Recall

🔧 Debug

expert

Why does this Isolation Forest model fail to detect anomalies?

You trained an Isolation Forest model but it flags almost all points as normal. What is the most likely cause?

ML Python

from sklearn.ensemble import IsolationForest
model = IsolationForest(contamination=0.1, max_samples=100)
model.fit(data)
preds = model.predict(data)
print(sum(preds == -1))  # Count anomalies

A max_samples is too high, causing overfitting

B The model needs more trees to detect anomalies

C The data is not scaled, causing model to fail

D The contamination parameter is set too low compared to actual anomaly rate