Analytics on large social networks
This repository contains assignment work from CSL7390-Social-Network-Analysis course.
- Structural Measures:
Analysis of Zachary's Karate Club dataset and ego-Facebook dataset.- Node Count, Edge Count, Average Degree
- Degree Distribution
- Triangles
- Diameter
- Connected Components
- Clustering Coefficient
- Betweenness centrality (from scratch)
- Closeness centrality (from scratch)
- K-cores (basic algorithm)
- Cliques (resource limits exceeded*)
- Network Models:
- Generation of random graph using custom function.
- Evaluation with benchmarks - networkx algorithms.
- Generation of scale-free graph.
- Structural Analysis of both the graphs for similar number of nodes and edges.
- Influence Maximization:
- Greedy algorithm over Independent Cascade model
- Greedy algorithm over Linear Threshold model
- Effectiveness of Greedy algorithm
- Dynamic Influence Maximization:
Studied different algorithms on dynamic influence maximization and found Sieve Streaming to be the most efficient algorithm providing similar accuracy as Greedy. Experiments were performed in the sliding window streaming model.- Greedy algorithm [Paper]
- Influential Checkpoints [Paper]
- Sieve Streaming [Paper]
- Random (Reservoir Sampling)
