Graphlab user guide笔记

https://dato.com/learn/userguide/index.html

Introduction

GraphLab Create is a Python package thatallows programmers to perform end-to-end large-scale data analysis and dataproduct development.

• Data ingestion and cleaning with SFrames. SFrame is an efficient disk-based tabular data structure that is not limited by RAM. This lets you scale your analysis and data processing to handle terabytes of data, even on your laptop.

• Data exploration and visualization with GraphLab Canvas. GraphLab Canvas is a browser-based interactive GUI that allows you to explore tabular data, summary plots and statistics.

• Network analysis with SGraph. SGraph is a disk-based graph data structure that stores vertices and edges in SFrames.

• Predictive model development with machine learning toolkits. GraphLab Create includes several toolkits for quick prototyping with fast, scalable algorithms.

• Production automation with data pipelines. Data pipelines allow you to assemble reusable code tasks into jobs and automatically run them on common execution environments (e.g. Amazon Web Services, Hadoop).

In this User Guide, you will learn how to use GraphLab Create to:

• munge and explore both structured and unstructured data
• use advanced machine learning methods to build predictive models and recommender systems
• put your code into production and use it for real-world applications

Working with data

Data science often requires manipulating data so that it is ready for analysis. This section describes how to

• import data
• reshape unstructured data
• make transformations of existing data quickly
• compute summary statistics
• easily visualize your data

Graph data

Graphs allow us to understand complex networks by focusing on relationshipsbetween pairs of items. Each item is represented by avertex in the graph, andrelationships between items are represented by edges.

To facilitate graph-oriented data analysis, GraphLab Create offers aSGraph object, a scalable graph data structure backed by SFrames.In this chapter, we show that SGraphs allow arbitrary dictionary attributes onvertices and edges, flexible vertex and edge query functions, andseamless transformation to and from SFrames.

Creating an SGraph

There are several ways to create an SGraph. The simplest is to start with anempty graph, then add vertices and edges in the form of lists ofgraphlab.Vertexandgraphlab.Edge objects. SGraphs are structuallyimmutable; in the following snippet, add_vertices andadd_edges both returna new graph.