def force_directed(ndf, ldf, iterations=1000):
## add force-directed layout coordinate
bn.add_force_directed_layout(ndf,
linksdf=ldf,
nw=None,
iterations=iterations)
return ndf
def tsne_layout(ndf, ldf, plot=True):
## add tsne-layout coordinates and draw
bn.add_layout(ndf, linksdf=ldf, nw=None)
ndf.rename(columns={"x": "x_tsne", "y": "y_tsne"}, inplace=True)
if plot:
bn.draw_network_categorical(nw,
ndf,
node_attr="keyword_theme",
x='x_tsne',
y='y_tsne',
plotfile="network_tsne.pdf")
return ndf
def force_directed(ndf, ldf, iterations=100, plot=True):
## add force-directed layout coordinates and draw
bn.add_force_directed_layout(ndf,
linksdf=ldf,
nw=None,
iterations=iterations)
if plot:
bn.draw_network_categorical(nw,
ndf,
node_attr="keyword_theme",
x='x_force_directed',
y='y_force_directed',
plotfile="network_fd.pdf")
return ndf
def spring_layout(ndf, ldf, iterations=1000):
print("Running spring Layout")
nw = bn.buildNetworkX(ldf)
# remove isolated nodes and clusters for layout
giant_component_nodes = max(nx.connected_components(nw), key=len)
giant_component = nw.subgraph(giant_component_nodes)
layout = nx.spring_layout(
giant_component, k=0.2, weight='weight',
iterations=iterations) # k is spacing 0-1, default 0.1
x = {n: layout[n][0] for n in giant_component.nodes()}
y = {n: layout[n][1] for n in giant_component.nodes()}
ndf['x_spring'] = ndf['id'].map(x)
ndf['y_spring'] = ndf['id'].map(y)
# place all disconnected nodes at 0,0
ndf['x_spring'].fillna(0)
ndf['y_spring'].fillna(0)
return ndf
def build_network(df, attr, blacklist=[], idf=True, linksPer=3, minTags=1):
'''
Run basic 'build tag network' without any plotting or layouts or file outputs.
Resulting network can then be enriched and decorated before writing final files.
df = nodes dataframe
attr = tag attribute to use for linking
blacklist = tags to blacklist from linking
linksPer = avg links per node
minTags = exclude any nodes with fewer than min Tags
Returns: nodes and links dataframes
Dependency: https://github.com/foodwebster/Tag2Network
'''
print("\nBuild Network")
df[attr] = df[attr].fillna("")
df = df[df[attr] !=
''] # remove any recipients which have no tags for linking
df = df.reset_index(drop=True)
taglist = attr + "_list" # name new column of tag attribute converted into list
df[taglist] = df[attr].apply(
lambda x: x.split('|')) # convert tag string to list
df[taglist] = df[taglist].apply(
lambda x: [s for s in x if s not in blacklist
]) # only keep keywords not in blacklist
df[attr] = df[taglist].apply(lambda x: "|".join(
x)) # re-join tags into string with blacklist removed
ndf, ldf = bn.buildTagNetwork(
df,
tagAttr=taglist,
dropCols=[],
outname=None,
nodesname=None,
edgesname=None,
plotfile=None, #str(networkpath/'RecipientNetwork.pdf'),
idf=idf,
toFile=False,
doLayout=False,
linksPer=linksPer,
minTags=1)
return ndf, ldf
def plot_network(ndf,
edf,
plot_name,
x='x_tsne',
y='y_tsne',
colorBy='Cluster',
sizeBy='ClusterCentrality',
sizeScale=100):
# draw network colored by creative style and save image
# ndf = nodes dataframe
# ldf = links dataframe
# plotname = name of file to save image (pdf)
# dependency: https://github.com/foodwebster/Tag2Network
nw = bn.buildNetworkX(edf) # build networkX graph object
node_sizes = ndf.loc[:, sizeBy] * sizeScale
node_sizes_array = node_sizes.values # convert sizeBy col to array for sizing
dn.draw_network_categorical(nw,
ndf,
node_attr=colorBy,
plotfile=plot_name,
x=x,
y=y,
node_size=node_sizes_array)
def tsne_layout(ndf, ldf):
## add tsne-layout coordinates and draw
# dependency: https://github.com/foodwebster/Tag2Network
bn.add_layout(ndf, linksdf=ldf, nw=None)
ndf.rename(columns={"x": "x_tsne", "y": "y_tsne"}, inplace=True)
return ndf
#from Network.BuildNetwork import addLouvainClusters
#from Network.ClusteringProperties import basicClusteringProperties
pd.set_option('display.expand_frame_repr',
False) # expand display of data columns if screen is wide
wd = pl.Path.cwd()
datapath = wd / "raw_data" / "10_yrs_of_TED_Network.xlsx"
resultspath = wd / "processed_data"
## read network file
ndf = pd.read_excel(datapath, sheet_name="Nodes")
ldf = pd.read_excel(datapath, sheet_name="Links")
# create networkx object
nw = bn.buildNetworkX(ldf)
def force_directed(ndf, ldf, iterations=100, plot=True):
## add force-directed layout coordinates and draw
bn.add_force_directed_layout(ndf,
linksdf=ldf,
nw=None,
iterations=iterations)
if plot:
bn.draw_network_categorical(nw,
ndf,
node_attr="keyword_theme",
x='x_force_directed',
y='y_force_directed',
plotfile="network_fd.pdf")