Complex Data Types (Advanced Data Mining Methods)

• Sequence data, time series
• Graph data, social networks
• Web data, data fusion
• Research frontiers

Sequence Data

You have an ordered list or stream, with or without time.

EX: biological sequence, stock price, transaction history

• Sequential pattern mining (frequent sub-sequence)
• Pairwise sequence alignment (ex BLAST for biological sequences)
• Sequence modeling/prediction (ex Markov chain model)

Time Series Data

Sequence data that changes over time.

Typically look at a combined signal of:

• T: overall trend
• C: cyclic patterns
• R: random noise
• A: anomalies

Graph Data

G = (V, E) relationships between entities, very powerful abstraction.

• find frequent sub graphs
• anomaly detection
• graph modeling
• link prediction

Online Social Network (OSN) Data

Special type of graph with users, groups, content, and interactions.

• OSN modeling and community detection
• topic modeling, sentiment analysis
• information diffusion, recommendation
• cyber-safety: malicious accounts and behaviors

Web Data

• use all methods: text, image, audio, video, links

Build search engines, recommender systems, knowledge graphs

Relates to data fusion, where you cluster multi modal data.

Check out KDD conference for state of the art.