1.2.2017 Data-driven business Professor Henri Schildt Graphs and networks 1.2.2017 Data-driven business Professor Henri Schildt
The relational way of thinking Sciences have traditionally focused on the attributes of various ”things” Network perspective draws focus on the relations in-between Examples Communication networks, information flow Friendships, interpersonal trust Flows of resources between companies Source e.g.: Emirbayer & Mische (1998), American Journal of Sociology
Network Analysis is: An alternative way to view the world Instead of the properties of actors, we examine the properties of relations among actors Selection of methodological tools Measures, algorithms, survey designs, data representation techniques, etc. Research program(s) in social sciences Directs the attention of scholars to certain research questions and constructs
The History of Social Network Analysis Morenon graafit piirrettiin käsin. Laatikot poikia, ympyrät tyttöjä. Moreno’s (1930) description of friendship ties in a school class. Data is based on self-reported information (a survey). Boxes are boys, circles are girls. ’M’ denotes mutual (bi-directional) friendship.
Network Analysis Timeline 18th century: graph theory Invented by Euler to solve the bridges of Köningsberg puzzle 1930s: Moreno invents sociographs 1950s: Harvard studies Hawthorne studies and ”Yankee city” 1970s: Methodological development ”Social Network” Journal founded in 1978 1990s: Social network analysis becomes common in strategic management & organization theory Physicists become increasingly interested in networks 2000s: ”Social networks” start-up hype Physicists re-invent many social network concepts
Attribute vs. relational data Networks are represented as graphs. Circles denote the members of the networks, nodes. The lines represent ties or relations between nodes. The radius of circles typically represents some quantitative attribute of the node. Width of lines represents tie strength. Motorola Attribute data corresponds to the properties of nodes. E.g. for comapnies: the number of employees, sales, revenue, profitability, market share, etc. 1 3 5 10 6 Relational data corresponds to ties (relations, connections, linkages) between nodes. In strategic networks, such data may include e.g.: the number of alliance deals, the extent of mutual trust, the intensity of communication, mutual investments, the amount of trade, and other forms of interdependence. 1 5 3 Nokia IBM Samsung Symbian
Relational and structural theories Network studies distinguish different forms of effects concerning (1) The relationship between two actors (2) The wider structure of the network Motorola Let’s suppose that ties represent knowledge flows between companies. What benefits does IBM gain from the fact that none of the other companies have ties to Samsung? Nokia IBM Samsung Symbian
Some relevant forms of networks Social networks Nodes are individuals Ties represent friendship, trust, communication intensity, task relatedness (workflow), etc. Strategic networks Nodes are firms Ties represent strategic alliances or other repeated interaction among companies Other interfirm networks Ties represent firm similarity, interdependence, flow of money, communication, knowledge flow, common members of the board
Gephi – plotting networks It’s not great, but it is free and powerful Loads of add-ons you can install to do stuff Let’s see if it works Open up first the Jazz network Filter nodes & edges Go to Window => Filters Select Edges => Weight (& adjust) Select Topology => Degree Range (& adjust)
Deep structural holes & good ideas
From affinity matrix to network We can turn attribute data to similarity data Can you figure out how similar students are based on the courses they attend? Can you figure out how similar the courses are based on the students who attend? Data-driven business Strategy work Intl. Human Resource Mngmt. Mary 1 Paul Jack Kate
How to create affinity matrix? To 4x4 matrix with similarity of people, choose 4*4 area, write the formula =MMULT(<matrix>; TRANSPOSE(<matrix>)) + press shift+ctrl+enter To create 3x3 matrix with similarity of courses, choose 3*3 area, write the formula =MMULT(TRANSPOSE(<matrix>); <matrix>)
Entering data into Gephi Compose an affinity network based either on the courses you have taken or your hobbies Include some friends you know Copy the matrix you created, and paste it into a new worksheet You need to do Paste Special & choose “Values” so that you don’t paste the formulas Alternatively, form the new matrix into a new sheet Save the worksheet as a csv file; now you can open it in Gephi
Can you use this approach to map out intensity of competition among airlines?
Free web-based tools quite strong http://visjs.org/network_examples.html www.theyrule.net https://quid.com/quid-in-action#innovation/multiple-views