Overfitting: High Training Accuracy, Low Validation Accuracy
A team is training a binary classification model using Amazon SageMaker. They notice that the training accuracy is 99% but the test accuracy is only 70%. Which technique should they apply first to address this?
Quick Answer
A 29-point gap between training accuracy and test accuracy is the textbook signature of overfitting: the model isn't learning the general patterns that separate the two classes, it's memorizing quirks specific to the training examples it happened to see, which lets it score near-perfectly on data it's already memorized while performing much worse on data it hasn't. Regularization is the first-line fix because it directly discourages that memorization at the mechanism level — adding an L1 or L2 penalty to the loss function punishes overly large or overly complex weight values, which pushes the model toward simpler decision boundaries that are less likely to have fit noise specific to the training set. In SageMaker this is typically as simple as setting an l1 or l2 hyperparameter on a built-in algorithm, or adding dropout layers if working with a custom framework, making it a low-effort first thing to try before reaching for heavier interventions like collecting more training data or redesigning the model architecture. Any scenario describing training accuracy far exceeding test accuracy, rather than both being uniformly low, is describing overfitting specifically — a different problem from underfitting, which shows up as both accuracies being low together, and calls for a different set of fixes.
⚠ Common exam trap
AWS often tests the misconception that overfitting is solved by increasing data or model complexity, when in fact the first step should be regularization to penalize overly complex models.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Apply regularization
The high training accuracy (99%) paired with significantly lower test accuracy (70%) is a classic symptom of overfitting, where the model memorizes the training data instead of learning generalizable patterns. Regularization (Option B) is the first-line technique to combat overfitting by adding a penalty to the loss function (e.g., L1 or L2 regularization), which discourages overly complex decision boundaries. In Amazon SageMaker, this can be implemented via hyperparameters like `l1` or `l2` in built-in algorithms or by adding dropout layers in a custom framework.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Reduce training data
Why it's wrong here
Reducing data would likely increase overfitting.
- ✓
Apply regularization
Why this is correct
Regularization adds penalty for large weights, helping to reduce overfitting.
- ✗
Increase learning rate
Why it's wrong here
A higher learning rate may cause unstable training but does not directly address overfitting.
- ✗
Increase model complexity
Why it's wrong here
Increasing complexity would likely worsen overfitting.
Go deeper
Related to this question
About these practice questions
This AIF-C01 question is part of Courseiva's 619-question bank — original exam-style content with full explanations and wrong-answer analysis, never real exam questions or exam dumps. Learn why practice questions differ from exam dumps →
Same concept, more angles
1 more way this is tested on AIF-C01
These questions test the same concept from different angles. Work through them to make sure you can recognise it however the exam phrases it.
Variation 1. A team trained a deep learning model that achieves 99% accuracy on training data but only 70% on validation data. What is the most likely issue?
easy- A.Underfitting
- ✓ B.Overfitting
- C.Data leakage
- D.Feature scaling
Why B: The model performs exceptionally well on training data (99% accuracy) but significantly worse on validation data (70% accuracy). This large gap indicates the model has memorized the training data, including noise and irrelevant patterns, rather than learning generalizable features — a classic symptom of overfitting.
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This AIF-C01 practice question is part of Courseiva's free Amazon Web Services certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AIF-C01 exam.