社交网络分析入门指南：初创企业实用教程

3星 · 超过75%的资源需积分: 49 83 浏览量更新于2024-07-21 1 收藏 12MB PDF 举报

《社交网络分析：初创企业的实用指南》（Social Network Analysis for Startups）由 Maksim Tsvetovat 和 Alexander Kouznetsov 联合撰写，是一本专为初学者设计的社交网络数据分析入门教材，也适合那些对机器学习感兴趣的初级读者。本书深入浅出地讲解了社交网络分析的基本原理、方法和在商业环境中的实际应用，旨在帮助创业者和开发者更好地理解和利用这些数据来优化策略，提升竞争优势。本书的核心内容涵盖了以下几个方面： 1. **社交网络定义与理论基础**：首先介绍了社交网络的基本概念，包括节点、边、度、路径、社区结构等基本术语，帮助读者建立对社交网络模型的理解。 2. **数据收集与预处理**：书中详细讲解了如何从各种社交媒体平台、公开数据集或用户生成的内容中获取有效数据，并阐述了数据清洗、格式化和标准化的重要步骤。 3. **分析方法**：涵盖了各种社交网络分析工具和技术，如度中心性、接近中心性、聚类分析、网络可视化等，这些都是理解社交网络动态的关键手段。 4. **社交网络洞察**：探讨如何通过分析社交网络数据来揭示用户行为模式、影响力传播、意见领袖识别以及潜在的合作或竞争关系，以便企业定位目标市场和客户群体。 5. **案例研究与实战应用**：书中提供了多个真实世界的例子，展示如何将理论知识转化为实际操作，使读者能够将所学应用到初创企业的具体业务场景中。 6. **与机器学习的结合**：强调了社交网络分析与机器学习技术的融合，例如推荐系统、情感分析等，以提升数据分析的精度和深度。 7. **版权和版权声明**：最后，本书遵循版权法规定，指出该书在美国发行并保留所有权利，同时提供了在线版本和联系信息供读者获取更多资源。《社交网络分析：初创企业的实用指南》是一本实用性强、内容全面的教程，无论是想要踏入社交网络数据分析领域的新手，还是希望提升现有技能的企业家和开发者，都将从中受益匪浅。通过阅读这本书，读者可以掌握社交网络分析的基石，为自己的业务决策提供有力的数据支持。

In this book, we will analyze social media data. We will harvest data from Twitter

(Appendix A), from Facebook and LiveJournal. We will learn to recognize online com-

munities, and study the anatomy of a viral video and a flashmob.

However, I will show that SNA can be applied in many different ways. In this book,

we will look at the social media, but we will also look beyond social media. We shall

study the relationships between companies through investment networks and through

shared boards of directors. We will look inside an organization and discover how the

social network around the water cooler and lunchroom affects the company’s ability

to perform—and how a company could shoot itself in the foot by ignoring this. We

will look at campaign finance and discover how a single special-interest group can

control the outcome of an entire election. We will explore the world of terrorists, rev-

olutionaries, and radicals—from stories of the Khobar Towers bombing in 1998 and

the 9/11 attacks, to the recent uprising in Egypt. We shall look at the anatomy of fads

and trends—which are often mediated by Twitter and Facebook, but are offline phe-

nomena by nature.

I will show you that network data is everywhere—you just need to learn to recognize

and analyze it. And once you do, new insights and ideas shall follow.

Analyzing Relationships to Understand People and Groups

The science of Social Network Analysis (SNA) boils down to one central concept—our

relationships, taken together, define who we are and how we act. Our personality,

education, background, race, ethnicity—all interact with our pattern of relationships

and leave indelible marks on it. Thus, by observing and studying these patterns we can

answer many questions about our sociality.

What is a relationship? In an interpersonal context, it can be friendship, influence,

affection, trust—or conversely, dislike, conflict, or many other things.

Binary and Valued Relationships

Relationships can be binary or valued: “Max follows Alex on Twitter” is a binary rela-

tionship while “Max retweeted 4 tweets from Alex” is valued. In the Twitter world,

such relationships are easily quantified, but in the “softer” social world it’s very hard

to determine and quantify the quality of an interpersonal relationship.

A useful stand-in for strength of an interpersonal relationship is frequency of commu-

nication. Besides being objectively measurable, frequency of communication has been

found by scientists to reflect accurately on the emotional content, and amount of in-

fluence in a relationship. This would, of course, not be true in many contexts (and you,

my dear reader, are probably busy coming up with counterexamples right now)—but

in many cases, for the lack of better data, frequency of communication works.

2 | Chapter 1: Introduction

Symmetric and Asymmetric Relationships

It is easy to see that some relationships are asymmetric by nature. Teacher/student or

boss/employee roles presume a directionality of a relationship, and do not allow for a

symmetric tie back. Following on Twitter and LiveJournal is directional by definition

—but a follow-back tie can exist, thus symmetrizing the relationship

Other relationships are symmetric. Facebook friends and LinkedIn connections require

mutual confirmation—the software forces a symmetry even when the real human re-

lationship is asymmetric.

In the real world, friendships and romantic relationships are asymmetric, as much as

we would like them not to be that way. Hence, we struggle with unrequited love, one-

sided friendships and other delusions of popularity. Given good data, we can study

these phenomena using SNA—but such data would be very difficult to obtain and

subject to self-reporting and other biases.

Multimode Relationships

Finally, we should mention that relationships can exist between actors of different types

—Corporations employ People, Investors buy stock in Corporations, People possess

Information and Resources, and so on. All of these ties are described as bimodal or 2-

mode—we will discuss them in detail in Chapter 5.

From Relationships to Networks—More Than Meets the Eye

If a traditional quantitative sociologist or econometrician got a hold of social network

data, he would approach it as follows:

• What can we learn about the respondents? Is there any demographic data on them?

Age, race, religion, income, education, location, and so on—any qualitative or

quantitative variables that could possibly be measured.

• What kind of quantitative metrics can be derived from the network data? Most

likely, these metrics would include various forms of centrality (see Chapter 3).

• What kind of quantitative or qualitative outcomes can be measured? That is, the

ability to integrate into a society as a citizen, the likelihood of trying illegal drugs,

etc.

Then he would build a multivariate regression model, controlling for some variables

and linking some of the other variables to outcomes. This is a perfectly valid approach

—in fact, this approach still gets a lot of mileage in the SNA conferences.

One of the traditional applications of this approach is the notion of homophily (Greek,

meaning love of the similar)—or, loosely, the notion that “birds of a feather flock to-

gether.” For example, it has been proposed that people of a similar age are more likely

to be friends then people of different generations, or that people of the same race tend

From Relationships to Networks—More Than Meets the Eye | 3

to associate together. While some of these conjectures ring true, they do not account

for such things as the complexity of internal dynamics of social groupings between

black teenagers in one high school class—where age, race, economic background, mu-

sic preferences, etc., may all have a high degree of similarity, and yet emotions run the

full spectrum of human conditions from love to rage.

However, we can do something that is remarkably different and, I dare say, better.

The standard statistical approach has one assumption—that of event independence,

or a Poisson process.

In a Poisson process, every event can be treated as a completely

independent occurrence, with no relationship to other events. Thus, we can compute

probability of an event based on outside variables or characteristics—and as a result,

come up with a decent model of reality. Of course, events are not always independent

—in which case, Bayesian statistics can create chains of dependencies and compute the

probabilities of macro-outcomes. Details of this are beyond the scope of this book, but

can be found in a variety of sources.

In a social network, we intuitively know that there are no independent events. People

form friendships and acquaintances through introduction—i.e., A meets B because A

and B both know C. There are love triangles (which, in the view of homophily, should

become more like a free-love commune!). There are long-range ties to unlikely others,

on the basis of homophily.

†

And counter-examples multiply.

When we do SNA, we remove the independence assumption and treat all ties as

potentially dependent on each other. This makes traditional statistical methods (e.g.,

regressions, or Markov Models) mathematically intractable on all but most trivial

problems—but have no fear, we shall develop and deploy new methods that are just

as powerful. We shall talk more about homophily in the context of information diffu-

sion in “How Does Information Shape Networks (and Vice Versa)?” on page 116.

Social Networks vs. Link Analysis

Another cousin to SNA is Link Analysis (LI). Some of you may have used LI in business

intelligence or law enforcement work, or seen it on TV. “Without a Trace” uses link

analysis in every episode; “Numbers” and “Law and Order” resort to it on occasion.

Link analysis is in many ways similar to SNA—both talk about relationships in terms

of nodes and edges (Figure 1-1) and both try to derive the idea of who is more important

in a network by analyzing the whole network, not individual events.

* http://en.wikipedia.org/wiki/Poisson_process

† Yours truly is friends with both people on the radical left and radical right of the political spectrum, yet

experiences little cognitive dissonance.

4 | Chapter 1: Introduction

The Power of Informal Networks

Let us start discussion of social network analysis with a little story.

ACME Consulting was an old auditing shop. Founded in the 1960s, it was a family-

owned business for more then 25 years. The firm grew slowly and organically for many

years, and was the owner’s pride and joy. ACME’s main business was accounting and

auditing—serving a few select, reliable clients. In the mid-1990’s, ACME succumbed

to the lure of the information age, and added an IT department that built all of the

infrastructure needed for running a “21st century operation”—but the core of the

business was old-fashioned, with a nose-to-the-grindstone understanding of the pa-

perwork and the numbers.

However, all good things must come to an end, and the owner, nearing his retirement

years, decided to move to Boca Raton and embark on a new career of fly fishing and

being a full-time grandfather. He didn’t want to sell the business, and instead hired an

outside CEO to run the company while the family remained nominally in control.

The org chart in Figure 1-2 shows how the CEO (Conrad) saw his company. The au-

ditors and the IT groups formed their own divisions, of about 100 people each (the org

chart below is greatly simplified). Separate was a “secretarial pool”—a group whose

sole responsibility was dealing with mountains of paperwork that the firm handled for

clients. In a way, this was a vestige of the “Mad Men” world—some of the older auditors

still referred to the secretaries as “girls,” despite the fact that many of them were in their

50s. The departments bickered with each other over resources and requirements, prob-

lem resolution took far too long and required too many meetings, and cost the company

many billable hours.

Figure 1-2. ACME Consulting org chart (before reorganization)

This way of doing business didn’t sit well with CEO Conrad, a young graduate of a

prestigious business school. He wanted to build a modern, client-oriented, responsive

6 | Chapter 1: Introduction

剩余189页未读，继续阅读

qq_24623969

粉丝: 0
资源: 1

社交网络分析入门指南：初创企业实用教程

社交网络分析及案例详解_PDF电子书下载 带书签目录 完整版.pdf

方滨兴 在线社交网络分析 方滨兴

在线社交网络分析-方滨兴

python社交网络分析工具

社交网络分析技术发展背景

d3.js 社交网络分析

Web data: Amazon movie reviews社交网络分析

Web data: Amazon Fine Foods reviews数据集可以用于社交网络分析吗？

做一个基于社交网络分析的舆情分析实验

如何将社交网络分析与机器学习、深度学习结合

最新资源

社交网络分析及案例详解_PDF电子书下载带书签目录完整版.pdf

方滨兴在线社交网络分析方滨兴