# +
from plotly.subplots import make_subplots
import plotly.graph_objects as go
fig = make_subplots(rows=4, cols=1, subplot_titles=list(dict_g.keys()), shared_xaxes=True)
for i, k_v in enumerate(dict_g.items()):
k, v = k_v
v0, v1 = v
fig.add_trace(go.Bar(x=v1.index, y=v1.values, name="+", marker_color='#2ca02c' ),col=1, row=i+1)
fig.add_trace(go.Bar(x=v0.index, y=v0.values,name="-", marker_color='#d62728'),col=1, row=i+1)
fig.update_layout(title="Overall J' per user-type", height=800)
# -
import utils
utils.plotly_to_tfm(fig, 'timeseries-sum-js-types-2')
import datetime
dict_g = {}
for i,g in enumerate(join_df.groupby(by=['positive'])):
g_id, g = g
values_overall = g.groupby('d').value.sum()
values_overall.index.name = 'index'
for x in range((values_overall.index.max()-values_overall.index.min()).days):
date_try = values_overall.index.min() + datetime.timedelta(days=x)
if date_try not in values_overall.index:
values_overall.loc[date_try]=0
index_sorted = values_overall.sort_index()
bias_g = get_bias(all_profiles_mod)
unbiased_series = pd.Series(index_sorted.reset_index().apply(lambda x: x[1]/bias_g[x[0]], axis=1).values, index=index_sorted.index)
dict_g[g_id]= unbiased_series
1002255096854515714 in all_profiles.index
len(all_profiles)
import os
tls = os.listdir("../data/models")
import
# 911220797976580096
load_and_plot(np.random.choice(tls)[:-4])
fig = load_and_plot('1015264264892960769')
fig.update_layout(title='user_id=455018574', margin=dict(l=0, r=0, t=40, b=0), height=400)
utils.plotly_to_tfm(fig, 'real-case-1')
fig = load_and_plot('455018574')
fig.update_layout(title='Real case scenario', margin=dict(l=0, r=0, t=40, b=0), height=400)
utils.plotly_to_tfm(fig, 'real-case-1')
import importlib
importlib.reload(utils)
# lambda_2 te 2 probs
fig = load_and_plot('112894609')
fig.update_layout(title='Multi-λ example', margin=dict(l=0, r=0, t=40, b=0), height=400)
utils.plotly_to_tfm(fig, 'multi-lambda')
fig = load_and_plot('112894609')
# %%
all_profiles.type_profile.unique()
# %%
import plotly.express as px
fig = px.scatter(all_profiles[
all_profiles.type_profile.isin(['random-friend', 'random-follower'])
& (all_profiles.followers_count < 5000) &
(all_profiles.friends_count < 5000)],
x='followers_count',
y='friends_count',
color='type_profile')
# %%
fig['data'][0]['marker']['opacity'] = 0.1
fig['data'][1]['marker']['opacity'] = 0.1
# %%
utils.plotly_to_tfm(fig, 'random-types-bias')
# %%
import utils
# %%
len(all_profiles)
# %%
850396856551186432 in bad_index
# %%
fig.add_trace(go.Scatter(x=freq.index,
y=utils.expected_texts_bipoisson(model_switch, freq),
mode='lines',
marker_color="#639ffa",
name='expected #tweets switch',
opacity=0.4),
row=3,
col=1)
fig.add_trace(go.Bar(x=freq.index,
y=freq.values,
name='# tweets',
marker_color="#FFA15A"),
row=3,
col=1)
utils.plotly_to_tfm(fig, 'bayesian-switch-sigmoid')
# utils.plot_everything({'lambda_1':early_disaster_rate[1].numpy(), 'lambda_2':late_disaster_rate[1].numpy(), 'tau_samples':switchpoint[1].numpy(),'tau':np.array([pd.to_datetime(freq.index.values[int(t)]) for t in switchpoint[1].numpy()])}, freq)
fig = make_subplots(rows=4,
cols=1,
specs=[[{}], [{}], [{
"rowspan": 2
}], [None]])
fig.update_layout(barmode='overlay', height=600)
utils.plot_lambdas(model_sigmoid['lambda_1'],
model_sigmoid['lambda_2'],
fig=fig,
row=1,
col=1)
utils.plot_lambdas(model_switch['lambda_1'],
model_switch['lambda_2'],
ego_df['weight']=1
ego_df = ego_df.drop(columns=['attributes'])
d3fdgraph.plot_force_directed_graph(ego_df)
all_profiles.sample().iloc[0].ego_network
example_profile = all_profiles[all_profiles.type_profile=='journalist'].sample().iloc[0]
fig2 = utils.plot_directed(example_profile.ego_network, example_profile.screen_name)
fig2
e
fig = utils.plot_directed(example_profile.ego_network, example_profile.screen_name)
utils.plotly_to_tfm(fig2, 'networks-ego-example-5')
all_profiles['ego_transitivity']=all_profiles.ego_network.apply(lambda G: nx.transitivity(G) if not pd.isnull(G) else np.nan)
selected = ((all_profiles.edges_between_outliers+all_profiles.edges_reciprocal_ego)>0)
all_profiles.head()
has_network = -all_profiles.ego_network.isnull()#&all_profiles.ego_network.apply(lambda x: len(x)>0 if not pd.isnull(x) else False)
def sample_hist(fig, series, name, color, n=25, n_samples=100, **kwargs):
fig.add_trace(go.Histogram(x=np.zeros(4), marker_color=color, legendgroup=name,name=name, visible='legendonly', showlegend=True), **kwargs)
for i in range(n):
showlegend = False
# if i==0:
fig = px.bar(x=total_count[2:].index, y=total_count[2:].values/np.hstack([a,[a[-1]]]))
# total_count.plot()
fig = px.bar(x=total_count_2[2:].index, y=total_count_2[2:])
fig
fig.layout['xaxis']['title']['text']='Time'
fig.layout['yaxis']['title']['text']='Mean tweet distribution'
final_bias = np.hstack([a,[a[-1]]])
import utils
utils.plotly_to_tfm(fig, 'intro-overall-activity-2')
import plotly.graph_objects as go
fig = go.Figure()
colors = ['#636EFA','#EF553B','#14D09E', '#AB63FA']
fig.add_trace(go.Scatter(x=total_count_2[2:].index, y=total_count_2[2:].values, name='RT_ratio', marker_color=colors[2]))
# fig.add_trace(go.Scatter(x=text_count_2[2:].index, y=text_count_2[2:].values, name='text_ratio', marker_color=colors[1]))
utils.plotly_to_tfm(fig, 'intro-overall-ratio-rt')
# for g_id,g in all_profiles[-all_profiles.dict_levels.isnull()].groupby('type_profile'):
g = all_profiles[-all_profiles.dict_levels.isnull()]
starting = g.dict_levels.apply(lambda x: list(x.keys())[0])