AI0-001 AI Models and Data Engineering Practice Test 5 — 15 Questions

Question 1

A data analyst is cleaning a dataset and finds that 20% of the values for the 'age' column are missing. Which imputation method is most robust if the data is not normally distributed?

Accepted Answer

Median imputation. Median imputation is the most robust method for handling missing values in the 'age' column when the data is not normally distributed because the median is unaffected by outliers or skewness. Unlike the mean, which is sensitive to extreme values, the median provides a central tendency measure that better represents the typical value in non-normal distributions, preserving the dataset's integrity for downstream modeling.

Answer

Mean imputation

Answer

Mode imputation

Answer

Remove rows with missing values

Question 2

Which TWO techniques are commonly used for feature selection in machine learning? (Choose 2)

Accepted Answer

L1 regularization (Lasso). L1 regularization (Lasso) is correct because it adds a penalty equal to the absolute value of the coefficients to the loss function, which drives less important feature coefficients exactly to zero, effectively performing embedded feature selection. Recursive Feature Elimination (RFE) is correct because it is a wrapper method that repeatedly trains a model, ranks features by importance (e.g., coefficients or feature_importances_), removes the weakest feature(s), and recurses until the desired number of features remains. PCA is not a feature-selection technique but a dimensionality-reduction method that creates new uncorrelated components from the original features, so it does not select among them. SMOTE is a data-level technique for handling class imbalance by synthesizing minority-class samples, not for selecting features. Dropout is a neural-network regularization method that randomly deactivates neurons during training to reduce overfitting, and it does not perform feature selection.

Answer

Principal Component Analysis (PCA)

Answer

SMOTE

Answer

Dropout

Question 3

Which THREE are common data preprocessing steps in a machine learning pipeline? (Choose 3)

Accepted Answer

Encoding categorical variables. Encoding categorical variables (B) is a standard preprocessing step because algorithms require numeric input, so techniques like one-hot encoding or label encoding convert strings into numbers. Scaling numeric features (D) is also preprocessing, since methods such as standardization (z-score) or min-max normalization bring features to comparable ranges, which helps distance-based and gradient-based models. Handling missing values (E) is preprocessing too, using imputation (mean, median, mode) or deletion to ensure the dataset is complete before training. Hyperparameter tuning (A) and model evaluation (C) are not preprocessing; they occur later in the pipeline, during model selection/training and after training, respectively.

Answer

Hyperparameter tuning

Answer

Model evaluation

Question 4

Which TWO are best practices for versioning machine learning models? (Choose 2)

Accepted Answer

Tag each model with training date, hyperparameters, and performance metrics. Option B is correct because tagging each model with its training date, hyperparameters, and performance metrics creates an auditable lineage that lets teams reproduce results, compare candidates, and roll back to a known-good model when production metrics degrade. Option C is correct because placing model code and configuration under a version control system such as Git provides immutable commit history, branching, code review, and the ability to correlate a deployed artifact with the exact source revision that produced it. Together, B and C satisfy the core ML versioning requirements of reproducibility, traceability, and governance. Option A is wrong because reusing one model version across all deployments eliminates the ability to distinguish, roll back, or A/B test different models. Option D is wrong because storing only the final binary without metadata makes the model impossible to reproduce, audit, or troubleshoot. Option E is wrong because manually renaming files is error-prone, unauditable, and does not capture training context or enable automated deployment pipelines.

Answer

Use the same model version for all deployments

Answer

Store only the final model binary without metadata

Answer

Manually rename model files with version numbers

Question 5

A data engineer is designing a pipeline for a streaming data application that uses a machine learning model to detect anomalies in real time. Which TWO practices should the engineer implement to ensure data quality and model reliability?

Accepted Answer

Use a sliding window for feature computation. Option C is correct because a sliding window computes features over a continuously advancing time range, which is essential for real-time anomaly detection since it captures recent, temporally relevant data points while maintaining feature freshness as the stream evolves. Option D is correct because implementing data validation checks at the ingestion point catches malformed, missing, or out-of-range records before they reach the model, preventing garbage-in-garbage-out failures and preserving both data quality and model reliability in a streaming pipeline. Option A is not appropriate because batch processing in fixed intervals introduces latency and defeats the real-time requirement of the streaming application. Option B is not required for data quality or model reliability; storing all raw data indefinitely is a retention/archival decision and does not itself validate or improve streaming data. Option E is not ideal because a fixed 24-hour retraining schedule cannot adapt to concept drift or anomalies that emerge within the stream, and retraining cadence should be driven by drift detection rather than an arbitrary fixed interval.

Answer

Use batch processing to transform data in fixed intervals

Answer

Store all raw data indefinitely for future analysis

Answer

Retrain the model on a fixed schedule every 24 hours

Question 6

A large e-commerce company uses a recommendation system based on collaborative filtering. The system uses a matrix factorization model that is trained nightly on the entire user-item interaction history. Recently, the company launched a flash sale with thousands of new products. Users are reporting that the recommendations are not showing the new products, even for users who have purchased them during the sale. The data engineering team notices that the new products have very few interactions in the training data. The model's loss on the validation set has increased, and the recall@10 metric has dropped from 0.45 to 0.32. The team needs to improve the recommendation of new items without retraining the entire model from scratch every hour. Which approach should the team take?

Accepted Answer

Use a hybrid model that combines collaborative filtering with content-based features from product metadata. A hybrid model that combines collaborative filtering with content-based features (e.g., product metadata like category, price, or description) can recommend new products even with zero or very few user interactions. The content-based component leverages item attributes to compute similarity between new and existing items, enabling the system to surface new products without requiring extensive interaction history. This approach addresses the cold-start problem for new items while preserving the collaborative filtering signal for established items, and it does not require retraining the entire model from scratch every hour.

Answer

Retrain the model every hour to incorporate new interactions quickly

Answer

Remove the new products from the recommendation pool until they accumulate enough interactions

Answer

Increase the number of latent factors in the matrix factorization model

Question 7

A data scientist trains a deep learning model on a large dataset. The training loss decreases steadily but the validation loss starts increasing after 20 epochs. The scientist uses early stopping with patience=5. Which of the following is the MOST likely cause and best corrective action?

Accepted Answer

Model is overfitting; add dropout regularization.. The training loss decreasing while validation loss increasing after 20 epochs is a classic sign of overfitting, where the model memorizes training data noise instead of generalizing. Early stopping with patience=5 would halt training after 5 epochs of no validation improvement, but the root cause is overfitting. Adding dropout regularization randomly drops neurons during training, forcing the network to learn more robust features and reducing overfitting.

Answer

Training data is not representative; collect more data.

Answer

Model is underfitting; increase model capacity.

Answer

Learning rate too high; reduce learning rate.

Question 8

A company streams sensor data from IoT devices. The data arrives as JSON messages at high velocity. Which data pipeline architecture is BEST suited to handle this streaming data for near-real-time analytics?

Accepted Answer

Stream processing using Apache Kafka and Spark Streaming.. Apache Kafka acts as a distributed, fault-tolerant ingestion layer that can handle high-velocity JSON messages, while Spark Streaming processes the data in micro-batches for near-real-time analytics. This combination provides the low-latency, scalable pipeline required for streaming IoT sensor data, unlike batch or single-node approaches.

Answer

Batch processing using Hadoop MapReduce every 24 hours.

Answer

Batch processing using nightly ETL jobs.

Answer

Single-node database with periodic inserts.

Question 9

A dataset for a binary classification problem has 95% of samples in class "0" and 5% in class "1". The data scientist trains a logistic regression model and achieves 95% accuracy. Which metric should the scientist primarily use to evaluate model performance?

Accepted Answer

Precision, recall, and F1-score.. In a highly imbalanced dataset (95% class 0, 5% class 1), accuracy is misleading because a model can achieve 95% accuracy by simply predicting the majority class for all samples. Precision, recall, and F1-score provide a more nuanced view of performance on the minority class, which is typically the class of interest in binary classification problems. The F1-score, in particular, balances precision and recall, making it the primary metric for evaluating model effectiveness on imbalanced data.

Answer

R-squared.

Answer

Accuracy.

Answer

Mean squared error.

Question 10

An e-commerce company needs to update its recommendation model continuously as user preferences change. The model currently retrains from scratch every night, but the training time is too long. Which approach would reduce training time while keeping the model up-to-date?

Accepted Answer

Implement incremental learning using online gradient descent.. Incremental learning using online gradient descent updates the model parameters with each new data point or mini-batch, avoiding the need to retrain from scratch. This approach significantly reduces training time while continuously adapting to changing user preferences, making it ideal for real-time recommendation systems.

Answer

Use dimensionality reduction on features.

Answer

Switch to a simpler model.

Answer

Increase the batch size for retraining.

Question 11

A machine learning model for credit card fraud detection is deployed. The model's precision is 0.95 and recall is 0.60. The business cost of missing a fraud is very high. Which of the following should the team prioritize to reduce the number of false negatives?

Accepted Answer

Decrease the classification threshold.. Decreasing the classification threshold makes the model more sensitive, classifying more transactions as fraudulent. This increases recall (reducing false negatives) at the cost of precision. Given the high cost of missing fraud, lowering the threshold is the direct way to capture more true positives, even if it increases false positives.

Answer

Use a different model algorithm.

Answer

Add more features.

Answer

Increase the classification threshold.

Question 12

Which THREE practices are recommended for versioning machine learning models in a production environment?

Accepted Answer

Use a model registry like MLflow or DVC.. Option A is correct because a dedicated model registry such as MLflow or DVC is purpose-built for tracking model artifacts, versions, and lifecycle stages, which is the recommended practice for production ML versioning. Option B is correct because storing metadata like hyperparameters and the training data hash ties each model version to its exact training configuration and dataset, enabling reproducibility and auditability. Option C is correct because automating deployment based on version tags ensures that only approved, traceable model versions reach production and keeps deployment consistent with the registry. Option D is not recommended because Git is designed for source code and text, not large binary model files, which bloat repositories and lack proper artifact lineage. Option E is wrong because discarding older models destroys rollback capability, reproducibility, and compliance auditing, and storage savings do not justify that risk.

Answer

Use Git to version model binaries.

Answer

Keep only the latest model to save storage.

Question 13

Refer to the exhibit. A data scientist reviews the MLflow run for a Random Forest model on customer churn data. What is the most likely issue with this model?

Accepted Answer

The model is overfitting because there is a large gap between train and validation accuracy.. A large gap between training accuracy (e.g., 0.99) and validation accuracy (e.g., 0.82) indicates that the Random Forest model has memorized the training data but fails to generalize to unseen validation data. This is the classic symptom of overfitting, where the model captures noise rather than the underlying pattern. In MLflow, comparing train and validation metrics directly reveals this discrepancy.

Answer

The model is underfitting because training accuracy is too high.

Answer

The model is performing well because validation accuracy is above 0.8.

Answer

The model has a data leak because dataset version is v2.

Question 14

A financial institution is training a risk assessment model. The dataset includes customer credit scores, income, age, and past loan defaults. During feature engineering, a data engineer creates a new feature 'income_to_debt_ratio'. Which type of feature engineering technique is this?

Accepted Answer

Feature combination. 'income_to_debt_ratio' is created by combining two existing features (income and debt) into a single derived feature. This is a classic example of feature combination (also known as feature crossing or feature construction), where arithmetic operations or logical rules are applied to existing variables to generate new predictive signals. The goal is to capture interactions or relationships that the original features alone may not express linearly.

Answer

Feature encoding

Answer

Feature scaling

Answer

Feature selection

Question 15

A machine learning team is developing a model to predict server failure from telemetry data. They use a deep neural network with 3 hidden layers. After training, the model achieves 99% accuracy on training data but only 85% on validation data. Which technique should the team apply to reduce the generalization error?

Accepted Answer

Apply L2 regularization. The model exhibits high variance (overfitting) because it achieves 99% accuracy on training data but only 85% on validation data. L2 regularization (also known as weight decay) adds a penalty proportional to the squared magnitude of the weights to the loss function, which discourages the network from fitting noise in the training data and improves generalization. This directly reduces the gap between training and validation performance.

Answer

Increase the number of hidden layers

Answer

Increase the learning rate

Answer

Add more training data