Maritime Risk Intelligence
Early warning for port disruptions — from raw AIS vessel data to a live risk dashboard.
A maritime risk-intelligence system that turns large-scale AIS vessel data into early warnings of port disruption. A Spark pipeline ingests raw vessel records into structured Parquet datasets, a spatio-temporal fusion step combines vessel movement, port activity, and weather, and graph-based risk propagation estimates how a disruption at one port can cascade across the network — surfaced on a streaming risk-monitoring dashboard.
Port disruptions ripple through global shipping, but the early signals are buried in huge, noisy AIS vessel feeds — and a problem at one port rarely stays local.
PySpark ingests AIS records into Parquet; vessel movement, port activity, and weather are fused spatio-temporally to flag abnormal port behaviour; NetworkX models the interconnected ports to propagate risk; and a Streamlit dashboard streams anomaly scores with EWMA smoothing, tracking MTTD and Risk Delta to measure how early disruptions are caught.
- 1Processing large-scale AIS vessel data into analysis-ready Parquet datasets.
- 2Fusing vessel, port, and weather signals across space and time.
- 3Modelling how a disruption at one port cascades across the network.
- PySpark pipeline turning raw AIS vessel records into structured Parquet datasets.
- Spatio-temporal fusion of vessel movement, port activity, and weather to flag abnormal port behaviour.
- Graph-based risk propagation (NetworkX) estimating how disruptions cascade across ports.
- Streaming dashboard with anomaly scoring, EWMA smoothing, and MTTD / Risk Delta metrics.