AWS Certified Machine Learning – Specialty (MLS-C01) retired on 31 March 2026. These questions remain useful AWS machine-learning study material, but they no longer map to a bookable exam. Candidates certifying today should look at AWS Certified Machine Learning Engineer – Associate or AWS Certified Generative AI Developer – Professional.
8 of the 31 questions in this set, with the correct answer marked and explained.
1. A machine learning team has several large CSV datasets in Amazon S3. Historically, models built with the Amazon SageMaker Linear Learner algorithm have taken hours to train on similar-sized datasets. The team’s leaders need to accelerate the training process. What can a machine learning specialist do to address this concern?
✗Use Amazon Machine Learning to train the models.
✓Use Amazon SageMaker Pipe mode.
Amazon SageMaker Pipe mode streams the data directly to the container, which improves the performance of training jobs. (Refer to this link for supporting information.) In Pipe mode, your training job streams data directly from Amazon S3. Streaming can provide faster start times for training jobs and better throughput. With Pipe mode, you also reduce the size of the Amazon EBS volumes for your training instances. B would not apply in this scenario. C is a streaming ingestion solution, but is not applicable in this scenario. D transforms the data structure.
✗Use AWS Glue to transform the CSV dataset to the JSON format.
✗Use Amazon Kinesis to stream the data to Amazon SageMaker.
2. A term frequency–inverse document frequency (tf–idf) matrix using both unigrams and bigrams is built from a text corpus consisting of the following two sentences:
1. Please call the number below.
2. Please do not call us. What are the dimensions of the tf–idf matrix?
✗(8, 10)
✗(2, 8)
✗(2, 10)
✓(2, 16)
There are 2 sentences, 8 unique unigrams, and 8 unique bigrams, so the result would be (2,16). The phrases are “Please call the number below” and “Please do not call us.” Each word individually (unigram) is “Please,” “call,” ”the,” ”number,” “below,” “do,” “not,” and “us.” The unique bigrams are “Please call,” “call the,” ”the number,” “number below,” “Please do,” “do not,” “not call,” and “call us.” The tf–idf vectorizer is described at this link.
3. A data scientist is working on optimizing a model during the training process by varying multiple parameters. The data scientist observes that, during multiple runs with identical parameters, the loss function converges to different, yet stable, values. What should the data scientist do to improve the training process?
✗Do not change the learning rate. Increase the batch size.
✗Increase the learning rate. Keep the batch size the same.
✗Keep the batch size the same. Decrease the learning rate.
✓Reduce the batch size. Decrease the learning rate
It is most likely that the loss function is very curvy and has multiple local minima where the training is getting stuck. Decreasing the batch size would help the data scientist stochastically get out of the local minima saddles. Decreasing the learning rate would prevent overshooting the global loss function minimum. Refer to the paper at this link for an explanation.
4. A data scientist is evaluating different binary classification models. A false positive result is 5 times more expensive (from a business perspective) than a false negative result. The models should be evaluated based on the following criteria:
1) Must have a recall rate of at least 80%
2) Must have a false positive rate of 10% or less
Must minimize business costs After creating each binary classification model, the data scientist generates the corresponding confusion matrix. Which confusion matrix represents the model that satisfies the requirements?
✗TN = 96, FP = 4 FN = 10, TP = 90
✓TN = 98, FP = 2 FN = 18, TP = 82
The following calculations are required: TP = True Positive FP = False Positive FN = False Negative TN = True Negative FN = False Negative Recall = TP / (TP + FN) False Positive Rate (FPR) = FP / (FP + TN) Cost = 5 * FP + FN A B C D Recall 78 / (78 + 22) = 0.78 79 / (79 + 21) = 0.79 90 / (90 + 10) = 0.9 82 / (82 + 18) = 0.82 False Positive Rate 9 / (9 + 91) = 0.09 1 / (1 + 99) = 0.01 4 / (4 + 96) = 0.04 2 / (2 + 98) = 0.02 Costs 5 * 9 + 22 = 67 5 * 1 + 21 = 26 5 * 4 + 10 = 30 5 * 2 + 18 = 28 Options C and D have a recall greater than 80% and an FPR less than 10%, but D is the most cost effective. For supporting information, refer to this link.
✗TN = 91, FP = 9 FN = 22, TP = 78
✗TN = 99, FP = 1 FN = 21, TP = 79
5. When evaluating a model after the training and testing process, you notice that the error rate during training is high but the error rate during testing is low. Which of the following could be the reason for obtaining these error rates?
✓You have a programmatic issue with your algorithm.
When training error is high and testing error is low, this is highly unusual as it infers that the model is somehow predicting better than the data which was used to train the model. This is usually an indicator of a data issue or some systemic problem in the algorithm.
✓You have a data issue with both your training and testing datasets.
✗You need to re-evaluate the section of your algorithm.
✗You should train for a longer period of time.
6. You are consulting for a logistics company who wants to implement a very specific algorithm for warehouse storage optimization. The algorithm is not part of the currently available SageMaker built-in algorithms. What are your options?
✗Post an incendiary message to Twitter hoping to shame AWS into adopting the specialized algorithm.
✓Build the algorithm in a docker container and use that custom algorithm for training and inference in SageMaker.
If SageMaker does not support a desired algorithm, you can either bring your own or buy/subscribe to an algorithm from the AWS Marketplace.
✗Wait until the algorithm is available in SageMaker before further work.
✓Search the AWS Marketplace for the algorithm. If it exists, deploy it using SageMaker for inferences.
7. You work as a data scientist for a small team in charge of collecting and processing data from handheld devices that air conditioning technicians use in the field. You've successfully set up consumer applications to consume, partition, compress, and store the data in Amazon S3. For environmental compliance, your teams must process the device data on a daily basis and send a report to management, who then forwards the results to the local government agencies. You have setup an AWS Glue Job that processes the data on a daily basis. The job ran fast and was low cost the first few weeks, but after several months of collecting data, you notice that the AWS Glue Job is taking longer to run and is costing more money. You also notice that each daily run is processing all the data, instead of only new data. Which of the following actions can you take to improve the speed and cost, and also disable the reprocessing of already processed data?
✗Disable the Job Bookmark feature.
✓Review the active executors, complete stages, and maximum needed executors in CloudWatch during previous job executions.
Enabling Job Bookmarks keeps track of previously processed data and processes new data since the last checkpoint. You can use job metrics in AWS Glue to estimate the number of data processing units (DPUs) that can be used to scale out an AWS Glue job. The number of maximum needed executors is computed by adding the total number of running tasks and pending tasks, and dividing by the tasks per executor. This result is a measure of the total number of executors required to satisfy the current load. Number of actively running executors In contrast, the number of actively running executors measures how many executors are running active Apache Spark tasks. As the job progresses, the maximum needed executors can change and typically goes down towards the end of the job as the pending task queue diminishes.
✓Enable the Job Bookmark feature.
✓Determine the optimal DPU capacity.
8. You want to design a model that will predict the price of a used car based on attributes of the car as a linear regression model. Which of the following algorithms could you use for your car price prediction model?
✓XGBoost
Linear Learner and XGBoost are two algorithms most closely identified with the use case of a linear regression problem.
✓Linear Learner
✗BlazingText
✗NTM
23 more questions in the app
Practise the full 31-question set with a timer, scoring and progress tracking.