SageMaker Canvas: No-Code Time Series Forecasting

Amazon SageMaker Canvas empowers users of all technical levels to perform time series forecasting without coding. Leveraging SageMaker Data Wrangler, it offers a no-code solution for data preparation, a crucial step often requiring expertise in statistical and data science methods. The process begins with data import from various sources, including local uploads, S3, Redshift, and various databases. Data Wrangler then provides automated recommendations and step-by-step guidance for data cleaning and transformation. Users can interact via a visual interface or use natural language prompts (‘Chat for data prep') to modify data, handle missing values, and convert data types. For time series forecasting, the system requires a timestamp column, a target column for forecasting, and an item ID column. Specific date formats are supported, and forecasting can be performed at various intervals (minute, hour, day, etc.). The prepared data is then seamlessly integrated into SageMaker Canvas for model building, offering both ‘Quick build' and ‘Standard build' options. The platform handles data security with features like Amazon EFS encryption for local imports and secure storage options like Amazon S3 with various encryption methods and access controls. While the no-code approach simplifies the process significantly, users still need to understand the fundamental requirements for time series data and model building. The platform's ease of use makes it ideal for business analysts, marketing professionals, and other users who lack coding expertise but need to perform accurate forecasting. The system's intuitive interface and natural language processing capabilities are key differentiators, enabling efficient data preparation and model training for time series forecasting.

(Source: https://aws.amazon.com/blogs/machine-learning/no-code-data-preparation-for-time-series-forecasting-using-amazon-sagemaker-canvas/)