SALESDECK2016v2

Social Data Science

DR PAUL SIEGEL/ DATA SCIENTIST

SALESDECK2016v2

You have a good sentiment model. Now what?

●  Identify influencers

●  Construct audiences

●  Activate advocates Key model: social networks

SALESDECK2016v2

NOW YOU KNOW | #NYKCONF

BRANDWATCH.COM

SALESDECK2016v2

Social Network Analysis

1.  Build network model from data

2.  Community detection

3.  Influence modelling / optimization

SALESDECK2016v2

Network Models

SALESDECK2016v2

Social graph: nodes are people, edges are: ●  Interactions

●  Followership

●  Latent influence

SALESDECK2016v2

Coupled Hidden Markov Models

SALESDECK2016v2

Community Detection

SALESDECK2016v2

Isoperimetric Problem

SALESDECK2016v2

Algorithms: ●  Fiedler vector

●  Louvain

●  Short random walks

SALESDECK2016v2

Short Random Walks

SALESDECK2016v2

Influence

SALESDECK2016v2

●  Linear threshold model

●  Independent cascade model

●  Centrality heuristics (degree, eigenvector, PageRank, etc.)

SALESDECK2016v2

Tech

SALESDECK2016v2

●  Networkx (Python)

●  Graphx / GraphFrames (Spark)

●  Graph databases?

SALESDECK2016v2

References

●  Everyone’s an Influencer: Quantifying Influence on Twitter by Watts et al ●  Modeling Temporal Activity Patterns in Dynamic Social Networks by

Vasanthan Raghavan et al ●  Four proofs for the Cheeger inequality and graph partition algorithms by

Chung ●  Maximizing the Spread of Influence through a Social Network by Kempe et

al ●  networkx.github.io ●  spark.apache.org/graphx