Oracle Cloud Infrastructure 2025 AI Foundations Associate Intermediate Practice Exam: Medium Difficulty 2025

A financial services company is building an AI system to detect fraudulent transactions. During testing, they discover that the model performs well on historical data but fails to identify new fraud patterns that emerge in production. Which AI concept best describes this problem, and what is the primary cause?

An e-commerce company wants to extract product information from supplier invoices in multiple formats (PDF, images, scanned documents). The solution needs to identify key fields like product names, quantities, and prices without extensive custom model training. Which OCI AI Service would be most appropriate?

A data scientist is evaluating a binary classification model for predicting customer churn. The model achieves 95% accuracy, but the business team reports that it's missing most of the actual churners. Upon investigation, only 3% of customers in the dataset actually churned. Which metric would better evaluate this model's performance, and why?

A retail company wants to implement a chatbot that can understand customer sentiment in product reviews and respond appropriately to complaints. The chatbot should also generate personalized product recommendations based on conversation context. Which combination of OCI AI Services would best address these requirements?

A healthcare organization is developing a machine learning model to predict patient readmission risk. They have features including age, diagnosis codes, previous hospitalizations, and socioeconomic data. During feature engineering, they discover strong correlations between multiple features. What problem does this present, and what is the most appropriate solution?

An organization wants to use a Large Language Model (LLM) to answer questions about their internal company policies and procedures. The information is confidential and frequently updated. They want accurate, up-to-date responses without exposing proprietary data for model retraining. Which approach would be most suitable?

A manufacturing company uses OCI Anomaly Detection to monitor equipment sensor data for predictive maintenance. The service flags several anomalies, but maintenance teams report many false positives. What approach should they take to improve the model's performance?

A data science team is preparing to train a supervised learning model. They split their dataset into training (70%), validation (15%), and test (15%) sets. After training, they iteratively adjust hyperparameters based on validation set performance until achieving optimal results, then evaluate on the test set. Why is this three-way split important?

A content moderation team wants to classify user-generated content into multiple categories simultaneously (e.g., a post could be both 'sports' and 'politics'). They're using OCI Language service for text classification. What type of classification problem is this, and what consideration is most important for model output interpretation?

A company is implementing a generative AI solution for automated report generation. They notice that the LLM occasionally produces plausible-sounding but factually incorrect information in the reports. What is this phenomenon called, and what technique can help mitigate it?