In this session, we show how graph technologies can be combined with machine learning techniques, and applied to real-world use cases, using medical, network security, and financial data.
00:00 – Welcome, the story so far. Jean briefly recaps previous sessions (graph intro, architecture)
06:20 – Benefits and main idea of using graphs for data analysis – Sungpack
09:00 – Approach 1: Classical method via graph algorithms. Sungpack shares an example of anomaly detection, using a medicare data set that’s publicly available. Anomalies in treatments provided by specific doctors are identified, which may indicate fraud.
23:40 -- Approach 2: Feeding graph data into a machine learning pipeline by generating features from graph algorithms. Sungpack walks through a security application example, where the problem is detecting malware in a network.
31:05 -- Zeppelin notebook based demo – malware detection
35:40 -- Approach 3: Graph embedding techniques (future directions). Sungpack discusses techniques now in the R&D phase, and possible use cases in life sciences and finance.
55:20 – Resources, Analytics and Data Summit (March 2019) event announcement, and wrapup
Slides are available here: https://www.slideshare.net/JeanIhm/when-graphs-meet-machine-learning
This is the 6th video in a series. Watch the previous sessions here: https://www.youtube.com/playlist?list=PL3ZqpALcm8HNVrhEE2j6m-hTCv4zbdFwa
Sign up for future AskTOM graph sessions: https://devgym.oracle.com/pls/apex/dg/office_hours/3084