#Start to import pandas, and numpy as we will need to use them for the importing.
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
#As the atrributes are as a .csv file, i will impport them seperately.
attributes = pd.read_csv("/work/Krack-High-Tec-Attributes.csv")
#The last three files to import is .txt-files, as I will import them seperately.
advice = pd.read_csv("/work/Krack-High-Tec-edgelist-Advice.txt", sep="\s+", header=None)
friendship = pd.read_csv("/work/Krack-High-Tec-edgelist-Friendship.txt", sep='\s+', header=None)
reports = pd.read_csv("/work/Krack-High-Tec-edgelist-ReportsTo.txt", sep='\s+', header=None)
#As the numbers can be quite confusing, I renamed the coloumns to source, target and connection.
advice = advice.rename({0: "origin", 1: 'target', 2 : "connection"}, axis=1)
friendship = friendship.rename({0: 'origin', 1: 'target', 2 : "connection"}, axis=1)
reports = reports.rename({0: 'origin', 1: 'target', 2 : "connection"}, axis=1)
#Just double check that everything is in order
reports.head()
!pip install networkx
import networkx as nx
1. Create a network
In the beginning I will create the edgelist and work a bit with them, before I move on to the node list.
Generate network objects for the companies organizational structure (reports to), friendship, advice
#Before I start to make a network, i will remove all off those rows,
#where there is no connection
advice_removedzero = advice[advice['connection'] !=0]
friendship_removedzero = friendship[friendship['connection'] !=0]
reports_removedzero = reports[reports['connection'] !=0]
#Afterwards we remove the "connection" column as we do not use that column anymore
advice_cleaned = advice_removedzero.drop("connection", axis = 1)
friendship_cleaned = friendship_removedzero.drop("connection", axis = 1)
reports_cleaned = reports_removedzero.drop("connection", axis = 1)
#Afterwards we do create the edgelists
edge_list_advice = list(zip(advice_cleaned["origin"], advice_cleaned["target"]))
edge_list_friendship = list(zip(friendship_cleaned["origin"], friendship_cleaned["target"]))
edge_list_reports = list(zip(reports_cleaned["origin"], reports_cleaned["target"]))
#Create an empty graph
G = nx.Graph()
#add the edges to the graph
G.add_edges_from(edge_list_advice)
G.add_edges_from(edge_list_friendship)
G.add_edges_from(edge_list_reports)
len(G.edges())
#So to show the relations in the organisations, I have created crosstabs for all of the edgelists
advice_cros = pd.crosstab(advice_cleaned.origin, advice_cleaned.target)
friendship_cros = pd.crosstab(friendship_cleaned.origin, friendship_cleaned.target)
reports_cros = pd.crosstab(reports_cleaned.origin, reports_cleaned.target)
#Show the advice crosstab
advice_cros
#Show the friendship crosstab
friendship_cros
#Show the report crosstab
reports_cros
from scipy import sparse
sparse_advice = sparse.csr_matrix(advice_cros)
sparse_friendship = sparse.csr_matrix(friendship_cros)
sparse_reports = sparse.csr_matrix(reports_cros)
print(sparse_advice)
print(sparse_friendship)
print(sparse_reports)
So as I have worked and figured a bit out between the edgelists, I will continue to work on my nodelist.
attributes
#We create a DataFrame of the given nodelist, and rename it to nodelist
nodelist = pd.DataFrame(attributes)
print(nodelist)
nodelist
#Change the format, so it does go by the number
node_attr = nodelist.set_index("ID").to_dict("index")
#Double check
node_attr
#Set the nodes
nx.set_node_attributes(G, node_attr)
#See the nodes
G.nodes(data=True)
#See the edges
G.edges()
#Chech the length of the nodes
len(G.nodes())
#Check the length of the edges
len(G.edges())
#Draw an ugly figure
plt.figure(figsize=(5,5))
nx.draw_kamada_kawai(G)
2. Analysis
A: Network level characteristics. Find the overal network level of: Density
#Find the density
density = nx.density(G)
print("Network density:", density)
A: Network level characteristics. Find the overal network level of: Transistivity (Clustering Coefficient)
#Find the triadic closure or ustering coefficient
triadic_closure = nx.transitivity(G)
print("Triadic closure:", triadic_closure)
A: Network level characteristics. Find the overal network level of: Reciprocity
reciprocity = nx.reciprocity(G)
print("Reciprocity", reciprocity)
Are relationships like friendship and advice giving usually reciprocal?
#So to answer the following, I will create two Gs that only includes respectively advice and friendship
#and still of course the same nodeset
G_advice = nx.Graph()
G_advice.add_edges_from(edge_list_advice)
nx.set_node_attributes(G_advice, node_attr)
plt.figure(figsize=(5,5))
nx.draw_kamada_kawai(G_advice)
reciprocity = nx.reciprocity(G_advice)
print("Reciprocity in advice", reciprocity)
G_friendship = nx.Graph()
G_friendship.add_edges_from(edge_list_friendship)
nx.set_node_attributes(G_friendship, node_attr)
plt.figure(figsize=(5,5))
nx.draw_kamada_kawai(G_friendship)
reciprocity = nx.reciprocity(G_friendship)
print("Reciprocity in advice", reciprocity)
Are friends of your friends also your friends? - No idea.
Are the employees generally more likely to be in a friendship or advice-seeking relationship?
len(G_advice.edges())
len(G_friendship.edges())
#As we can see of those results, the employees tend to seek advice rather than have friends.
Who is most popular in the networks. Who is the most wanted friend, and advice giver?
nx.degree_centrality(G_advice)
#So as we can sse, it is ID 15 which is the best advice giver
nx.degree_centrality(G_friendship)
#As we can see here, number 13 is the most popular guy, by having the most friends.
Are managers in higher hirarchy more popular as friend, and advice giver?