Complex networks in applied research

Transcription

1 IBM Research Ireland Smarter Urban Dynamics Complex networks in applied research Michele Berlingerio, PhD, Research Staff Member IBM Corporation

2 IBM Research: 5 new labs established since 2010

3 The eras of IBM Research 2015: The Cognitive Era Cognitive computing systems learn and interact naturally with people to extend what either humans or machine could do on their own. They help human experts make better decisions by penetrating the complexity of Big Data.

4 On (Big) Data Mining If you torture the data long enough, it will confess. [Attributed to Ronald Coase, economist] The Data Miner s view: The problem with confession is that it s easy to get answers, it s not easy to get the truth.

5 On (Big) Data Mining We don t make an hypothesis and test it against the data, we extract the hypothesis from the data, using frequency and/or similarities Finding frequent patterns Clustering data by similarities Finding outliers Finding temporal trends Extracted knowledge represents what happened in the data (descriptive analytics) Results can be used to make predictions (predictive analytics) Supports decision making (prescriptive analytics)

6 Multidimensional network analysis Definition: a network where there may exist multiple connections (dimensions) between any two nodes is multidimensional friends colleagues same name colleagues same team relatives

7 Planetary emergency: world s urbanization 54% of the world s population lives in cities today By 2050, it will be around 70% 1,2 Handling urban dynamics is crucial! Smarter water management Smarter energy management Smarter transport Safer cities html

8 Complex networks in Smarter Urban Dynamics: Public safety Challenge Solution Impact Identify and characterize public safety related incidents from social media: Detecting location of incidents Finding a description for them Assessing their social impact Local clustering of spatiotemporal and textual graph built from social media data. Automated framework to find geo-located incidents. Language independent. No need for a priori keyword selection and filtering. Example of result (left) and User Interface (right) Tested on data from: New York City, Kenya, Ireland

9 Crowd-steering to avoid congested tourist areas, and to promote less popular areas Gather historical mobility patterns Predict congestion Real-time monitoring and replanning Complex networks in Smarter Urban Dynamics: Smarter tourism Challenge Solution Impact Data mining on mobile phone data Robust activity planning on multi-modal urban transport network Mobile application with tracking and feedback Less pedestrian congestions Faster visits (30 mins avg) Less popular areas get more visibility 5% increase in local economy Example of problem (left) and User Interface (right) Tested on data from: Venice

10 Complex networks in Smarter Urban Dynamics: Sociable carpooling Challenge Solution Impact Increase adoption of carpooling Put together in a car people who would enjoy the ride together Finding enjoyable carpooling matches Build multidimensional network of friendship, interests, and spatiotemporal compatibility Multiobjective social-green optimization of cars Fewer cars in the system (56% reduction) More enjoyable cars (100% increase) Attracts more people to carpooling (39% increase) User survey (left) and visualized results (right) Tested on data from: Rome, San Francisco Bay Area

11 Challenges What can we learn and predict by using complex network analysis? How can we inject complex network mining in cognitive systems? Enable prescriptive analytics in cognitive systems for: Urban dynamics Financial networks Social networks

12 References Papers: M. Berlingerio, M. Coscia, F. Giannotti, A. Monreale, D. Pedreschi. Multidimensional networks: foundations of structural analysis. In World Wide Web Journal, doi: /s : M. Berlingerio, F. Pinelli, F. Calabrese. ABACUS: Apriori-BAsed Community discovery in multidimensional networks. In Data Mining and Knowledge Discovery 27(3): M. Berlingerio, F. Calabrese, G. Di Lorenzo, X. Dong, Y. Gkoufas, D. Mavroeidis. SaferCity: a System for Detecting and Analyzing Incidents from Social Media. In ICDM Demos R. Guidotti, A. Sassi, M. Berlingerio, A. Pascale, B. Ghaddar. Social or green? A data-driven approach for more enjoyable carpooling. In IEEE Intelligent Transportation Systems (ITSC):

13 References Media coverage: mila-le-idee html?ref=ig stories/2015/09/21/foto/innovazione_premiati_i_vincitori_del_big_data_challenge /1/\#

14 Thank you! ie.ibm.com