AutoGPT/docs/content/platform/blocks/update/sampling.md

Data Sampling

What it is

A versatile data sampling tool that can select specific items from a dataset using various selection methods.

What it does

Takes a collection of data items and returns a smaller subset based on user-defined criteria and sampling methods. It can work with different types of data collections and offers multiple ways to choose which items to include in the sample.

How it works

The system looks at your data collection and selects items based on your chosen sampling method. It can pick items:

• Completely randomly
• At regular intervals
• From specific groups proportionally
• Based on importance weights
• In clusters or groups
• From the beginning or end
• Using reservoir sampling for streaming data

Inputs

• Data: The collection of items you want to sample from
• Sample Size: How many items you want in your final selection
• Sampling Method: How you want to choose the items (random, systematic, top, bottom, stratified, weighted, reservoir, or cluster)
• Accumulate: Whether to collect data over time before sampling
• Random Seed: A number to ensure you get the same results each time (optional)
• Stratify Key: The category to use when ensuring balanced group representation
• Weight Key: The value to use when considering item importance
• Cluster Key: The group identifier for cluster-based sampling

Outputs

• Sampled Data: The selected items from your dataset
• Sample Indices: The positions of the selected items in the original dataset

Possible use cases

• Quality control in manufacturing: Randomly selecting products for inspection
• Market research: Selecting a representative group of customers to survey
• Data analysis: Creating balanced training datasets for machine learning
• Scientific research: Selecting specimens for detailed analysis
• Social studies: Choosing participants for a study while maintaining demographic balance