Simplifying Time Series Forecasting with Amazon SageMaker Canvas

Overview of Time Series Forecasting

Solution Overview

Step-by-Step Walkthrough

Prerequisites

Detailed Solution Walkthrough

Key Features of SageMaker Data Wrangler

Data Preparation Techniques

Handling Missing Values

Resampling Time Series Data

Model Training in SageMaker Canvas

Clean-Up and Maintenance

Conclusion

About the Author

Simplifying Time Series Forecasting with Amazon SageMaker Canvas

Time series forecasting is essential for businesses attempting to predict future trends based on historical data patterns. This practice is crucial across various sectors, serving purposes like sales projections, inventory management, and demand forecasting. However, traditional approaches have often required deep expertise in statistical methods and data science to process raw time series data effectively. Enter Amazon SageMaker Canvas and SageMaker Data Wrangler—a game-changing combination that democratizes access to advanced data preparation and modeling techniques.

Solution Overview

Amazon SageMaker Data Wrangler assists users in preparing data for predictive analytics without needing programming knowledge. This solution lays out the steps associated with leveraging SageMaker for time series forecasting, including:

• Data Import from Varying Sources: Support for multiple datasets ensures versatility.
• Automated No-Code Algorithmic Recommendations: Streamlines the data preparation process.
• Step-by-Step Processes for Preparation and Analysis: Guides users through the workflow.
• Visual Interfaces for Data Visualization and Analysis: Enhances understanding and insights.
• Export Capabilities Post Data Preparation: Facilitates easy sharing and further analysis.
• Built-in Security and Compliance Features: Ensures data integrity and safety.

In this post, we’ll dive deeper into using SageMaker Canvas for data preparation in time series forecasting.

Walkthrough

Ready to Start?

To follow this walkthrough, you’ll need access to Amazon SageMaker Canvas and a synthetic dataset related to consumer electronics. This dataset contains historical price data that corresponds to sales transactions, essential for improving prediction accuracy—especially in consumer electronics, where pricing changes can significantly affect buying behavior.

Prerequisites

Before diving in, ensure you have:

1. An AWS account.
2. Access to Amazon SageMaker.
3. The consumer_electronics.csv dataset.

Using SageMaker Canvas

1. Sign In to AWS Management Console
   Go to Amazon SageMaker Canvas. On the Get Started page, select the Import and Prepare option.

2. Import Your Dataset
   Choose Tabular Data, as this is necessary for time series forecasting. Here, you’ll have multiple options for importing your dataset, including:

   • Local upload
   • Canvas Datasets
   • Amazon S3
   • Databases like Amazon Redshift, MySQL, PostgreSQL, etc.

   For demonstration purposes, we will select the local upload option.

3. Configuration Settings
   Select the consumer_electronics.csv file, and use the Import Settings panel to set your configurations, default values are often sufficient for this demo.

4. Data Modification
   Once the import process is complete, you’ll want to clean up the data to prepare it for forecasting. SageMaker Canvas provides various options:

   • Chat for Data Prep: Ideal for users unfamiliar with technical jargon.
   • Add Transform: Best suited for data professionals who understand the transformations needed.

   For instance, you can remove dollar signs from your data by asking the chat, “Can you get rid of the $ in my data?” and it will generate the necessary code for you.

5. Handling Missing Data
   Identify missing data points and apply transformations. You can drop or infer missing values by selecting options from the Handle Missing transform.

6. Resampling the Data
   To ensure your data fits the forecasting needs, you might need to resample it based on frequency. Use the Time Series transform options within SageMaker Data Wrangler to set this up.

7. Validation and Model Creation
   Select Create Model and run validation to ensure everything has been set up correctly before proceeding.

Security and Compliance

Amazon SageMaker ensures that your data is stored securely. You can choose between Amazon EFS for temporary storage or Amazon S3 for long-term solutions, complete with robust security features such as AWS Identity and Access Management (IAM) roles for access control.

Clean Up

To avoid incurring unnecessary charges:

• Delete your SageMaker Data Wrangler data flow and any S3 buckets created.
• Navigate to the SageMaker console, find your data flow, and delete it.

If you used S3 for storage, open the Amazon S3 console, find the relevant bucket, and delete it.

Conclusion

Amazon SageMaker Data Wrangler delivers a no-code solution for time series data preparation, making it accessible to users irrespective of technical background. Through an intuitive interface and natural-language-powered tools, users can prepare their data efficiently for forecasting needs. This transformation not only saves time but also empowers a diverse range of professionals to rely on data-driven insights for decision-making.

About the Author

Muni T. Bondu is a Solutions Architect at Amazon Web Services (AWS), based in Austin, Texas. She holds a Bachelor of Science in Computer Science, specializing in Artificial Intelligence and Human-Computer Interaction, from the Georgia Institute of Technology.

By utilizing tools like Amazon SageMaker Canvas and Data Wrangler, businesses can eliminate traditional barriers to data forecasting, ensuring that everyone—from analysts to executives—can participate in data-driven strategies. Happy forecasting!