AIToday
Large Language ModelsQiita 機械学習Published: Nov 20, 2026, 22:01 JST

AWS AIプラクティショナー guide: pick AI field by input data

Send on LINE
AWS AIプラクティショナー guide: pick AI field by input data

3 Key Points

  1. What happened

    The guide maps input types to fields — images and video to computer vision (CV), text to natural language processing (NLP), and numeric changes over time to time-series forecasting — and pairs keywords like 外観検査 with Amazon Rekognition and 感情 with Amazon Comprehend.

  2. Why it matters

    It flags that Amazon Forecast ended new sign-ups on July 29, 2024 and Amazon Fraud Detector on November 7, 2025, so older study materials pointing to them appear outdated, with SageMaker Canvas and AutoGluon named as the current paths.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

The guide is the second part of a series on the AWS Certified AI Practitioner exam, following an earlier installment that covered the exam's overall shape and basic AI/ML terms. This part targets task 1.2, identifying practical AI use cases, and its central teaching device is a lookup table that links problem-statement keywords to a field and an AWS service. The author stresses that the starting point should be what goes in and what comes out, not the service name, since managed AI services are pre-trained for specific tasks and can be used without machine-learning expertise.

A recurring theme is that service names age badly. The guide notes that Amazon Forecast stopped accepting new customers on July 29, 2024 and that Amazon Fraud Detector did so on November 7, 2025, with AWS pointing users to SageMaker Canvas and AutoGluon respectively. That is why older materials that still feature Forecast or Fraud Detector are presented as no longer usable for new work.

The guide also draws boundaries between easily confused ideas: OCR reads characters from images and belongs to CV, while analyzing the meaning of that text is NLP; regression predicts a numeric value at a point in time, while time-series forecasting uses time-ordered history to predict the future; classification sorts into preset categories, while anomaly detection looks for patterns that differ from the norm. It closes by noting that conventional ML models may fit better than foundation models where regulation, explainability or operational constraints are strict, and that AI/ML is the wrong choice when every case needs an identical correct answer or when a simple rule is cost-effective.

FAQ
Which AWS service does the guide recommend for time-series forecasting now?
Amazon SageMaker Canvas, which lets users build time-series forecasting models without writing code. The older dedicated service Amazon Forecast stopped accepting new customers on July 29, 2024.
What replaced Amazon Fraud Detector for fraud detection?
The guide says AWS recommends the open-source AutoML library AutoGluon, with training done in AutoGluon and models deployed on SageMaker AI. Amazon Fraud Detector ended new sign-ups on November 7, 2025.
When should a business not use AI or ML at all?
When every case needs exactly the same correct answer, such as tax calculation or applying list prices, since ML predicts and sometimes errs. Also when a simple rule achieves nearly the same result and the cost is not justified.
Qiita 機械学習Read Original Article
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleRTX Spark laptop beats AMD Ryzen AI Halo in AI speed test, but LTT Labs says not worth the price