Zinnya DEL VILLAR & Christophe THOVEX

Transcription

1 Zinnya DEL VILLAR & Christophe THOVEX

2 Approaching Big Data from a business perspective

3 What is Big Data?

4 What is Big Data? - IoT - Internet - Unstructured - Semi-structured - Structured Volume Variety Big Data Velocity Veracity - Speed of generation - Rate of analysis - Untrusted - Uncleansed

5 Big Data eliminates intuition

6 Big Data eliminates intuition How? Decisions can be made with a structured approach, through data driven insight.

7 Big Data life cycle Creation

8 Big Data life cycle Creation Processing

9 Big Data life cycle Creation Processing Output

10 Big Data life cycle Creation Processing Output Ressources and processes

11 Big Data life cycle Ressources and processes Business Engineer Architect Engineer Data Scientist Geomaticien Network Scientist Dev DevOps

12 Big Data value chain Improve the efficiency and effectiveness of every decision and/or action

13 Big Data and analytics Continuous feedback loop Manage data Relevant data External data Perform analytics Insights Rules/ Algorithms Drive decision Advanced analytics Predictive Business Intelligence Descriptive analytics Prescriptive analytics Predictive analytics Mathematical complexity

14 Benefits and risks of Big Data Availability of data Compute/analyze Deliver/extract value - Governance - Management - Architecture - Usage - Quality - Security - Privacy Big Data Success

15 DATA PRODUCTS

16 Product improvement Help human decisions Increase the production performance FOOD INDUSTRY Logistic optimisation Understand customer behavior

17 Network Science The study of network representations of physical, biological, and social phenomena leading to predictive models of these phenomena [1] [1] Committee on Network Science for Future Army Applications (2006). Network Science. National Research Council. ISBN

18 Network Science and Big Data Processing flows through hundreds of thousands of ties

19 Network Science and Big Data Processing statistics and/or probabilities as weights/flows in Bayesian/Markovian networks, Convolutional Neural Networks (Deep Learning) Social and Semantic Networks (SSN)

20 Network Science and Big Data Large temporal graph as probabilistic chains for predicting users behaviour

21 Network Science and Big Data Community clustering and characterizing

22 Network Science and Big Data Producers/Consumers Leaders Community clustering and characterizing

23 Network Science and Big Data Producers/Consumers Leading items Leaders Strategic items Community clustering and characterizing

24 Stocks and sales by fish species multiscale visualization Focused harbour

25 Stocks and sales by fish species multiscale visualization Leading species in sales at national scale Focused harbour Local leading specie in sales at the focused harbour Business strength : focus Lorient

26 Hidden sentiment extraction from the talk of crowds Main topic

27 Hidden sentiment extraction from the talk of crowds Closest topics

28 Data Science & Networks Quantitative analysis : Descriptive statistics, inferential statistics Descriptors Estimators

29 Data Science & Networks Qualitative analysis : To uncover and understand the big picture, using the data to describe the phenomenon and what this means. Semantic and Social Web, Linked Data, Ontologies for information retrieval Internet is_a communication network = true Linked Data is_a data network = true Ontology is_a semantic network = true

30 Data Science & Networks Qualitative analysis : From linguistic statistics towards semantic inferences and fuzzy reasoning Axiom : Birds to fly

31 Data Science & Networks Qualitative analysis : From linguistic statistics towards semantic inferences and fuzzy reasoning Axiom : Birds to fly p (Birds to fly) = 0.99

32 Data Science & Networks Qualitative analysis : From linguistic statistics towards semantic inferences and fuzzy reasoning Axiom : Birds to fly Lexical ambiguities in semantic networks p (Birds to fly) = 0.99

33 Data Science & Networks Qualitative analysis : Fuzzy reasoning and Analytic Intelligence Axiom : Birds to fly Discrimination of lexical ambiguities in semantic networks p (Birds to fly) = 0.99 p (Fruit to fly) = 0.001

34 Big Data + Data Science + Network Science è From machine learning towards machine reasoning p (Kiwi is_bird) = 0.5 p (Kiwi is_fruit) = 0.5 p (Birds to fly) = 0.99 p (Kiwi to fly) = 0.01 p (Kiwi to fly) = p (Fruit to fly) =

35 Big Data + Data Science + Network Science è From machine learning towards machine reasoning p (Kiwi is_bird) = 0.5 p (Kiwi is_fruit) = 0.5 p (Birds to fly) = 0.99 p (Kiwi to fly) = 0.01 Kiwi(this) is_fruit Kiwi(this) is_bird p (Kiwi to fly) = p (Fruit to fly) =

36 Big Data Science & Networks Other Works : - Daily recommendations for high stock availability reducing distribution costs in round trips with stock return from delivery points and constrained transportation capacity. Perspective example : - Fuzzy reasoning on the Game Theory for Trading and Marketing

37 THANK YOU