def plot_model():
global model_data
# Getting the predicted data frame from predict crime method.
model_data = predict_crime()
df_final_plot = model_data
# Creaing labels for shanky diagram
df_final_plot['TYPE_NO'] = df_final_plot['TYPE']
df_final_plot['TYPE_NO'].replace('THEFT', '2', inplace=True)
df_final_plot['TYPE_NO'].replace('MISCHIEF', '0', inplace=True)
df_final_plot['TYPE_NO'].replace('BREAK And ENTER', '1', inplace=True)
df_final_plot['TYPE_NO'].replace('VEHICLE COLLISION', '3', inplace=True)
df_final_plot['TYPE_NO'] = df_final_plot['TYPE_NO'].astype('int32')
NEIGHBOURHOOD = go.parcats.Dimension(
values=df_final_plot.NEIGHBOURHOOD, label="NEIGHBOURHOOD")
DAY_TYPE = go.parcats.Dimension(values=df_final_plot.DAY_TYPE, label="DAY_TYPE")
TYPE = go.parcats.Dimension(values=df_final_plot.TYPE, label="CRIME TYPE")
color = df_final_plot.TYPE_NO;
# Creating the shanky diagram with plotly go.
fig = go.Figure(data=[go.Parcats(dimensions=[NEIGHBOURHOOD, DAY_TYPE, TYPE],
line={'color': color, 'colorscale': 'rdbu'},
labelfont={'size': 12, 'family': 'Times'},
tickfont={'size': 10, 'family': 'Times'},
arrangement='freeform')])
return fig
def PCVisualize(filenum, columns, color):
filepath = session.get('filepath', None)
set_vars(filepath)
set_mappings()
global data_cols,df_dataset
cols_file = 'static/parallel.csv'
if(filenum == '1'):
datadf = df_dataset
columndata = columns.split(',')
columndata = [data_cols[c] for c in columndata]
resultdf = datadf[columndata]
# resultdf['color'] = datadf[data_cols[color]]
try:
col_len = len(columndata)
dict_list = []
for a in range(col_len):
dict_list.append(dict(label=str(columndata[a]), values = resultdf[columndata[a]]))
# data = [go.Parcats(line = dict(color = datadf[data_cols[color]], colorscale = 'rainbow', showscale = True, cmin=datadf[data_cols[color]].min(), cmax=datadf[data_cols[color]].max()), dimensions = dict_list)]
data = [go.Parcats(line = dict(color = datadf[data_cols[color]], colorscale = 'viridis', showscale = True, cmid=datadf[data_cols[color]].median()), dimensions = dict_list)]
graphJSON = json.dumps(data, cls=plotly.utils.PlotlyJSONEncoder)
para = graphJSON
except Exception as e:
return render_template("500.html", error=str(e))
return render_template("paraplot.html", plot=para)
def update_figure():
fig = go.Figure(data=[
go.Parcats(dimensions=[Time_dim, Terminal_dim, Passengers_dim],
line={
'color': color,
'colorscale': colorscale
},
hoveron='color',
hoverinfo='count+probability',
labelfont={
'size': 18,
'family': 'Times'
},
tickfont={
'size': 16,
'family': 'Times'
},
arrangement='freeform')
])
return {fig}
def feature_interactions(radio, url, feat_importance, rows):
data_id = int(re.search(r"data/(\d+)", url).group(1))
if feat_importance == "done":
df = pd.read_pickle("cache/df" + str(data_id) + ".pkl")
fi = pd.read_pickle("cache/fi" + str(data_id) + ".pkl")
else:
return []
# Get meta data
meta_data = pd.DataFrame(rows)
try:
target_attribute = meta_data[meta_data["Target"] == "true"][
"Attribute"
].values[0]
target_type = meta_data[meta_data["Target"] == "true"]["DataType"].values[0]
except IndexError:
return "No target found", "No target found"
if target_type == "nominal" or target_type == "string":
y = pd.Categorical(df[target_attribute]).codes
else:
y = df[target_attribute]
# Feature interaction plots
df = clean_dataset(df)
# Extract top nominal, top numeric features
numerical_features = list(
meta_data["Attribute"][meta_data["DataType"] == "numeric"]
)
nominal_features = list(
meta_data["Attribute"][meta_data["DataType"] == "nominal"]
)
top_numericals = fi["index"][fi["index"].isin(numerical_features)][:4]
top_nominals = fi["index"][fi["index"].isin(nominal_features)][:4]
df["target"] = df[target_attribute]
# Bin numeric target
if target_type == "numeric":
# cmap_type = 'seq'
df["target_var"] = y
df = bin_numeric(df, "target_var", "target")
df.drop("bin", axis=1, inplace=True)
df.drop("target_var", axis=1)
else:
# cmap_type = 'cat'
try:
df["target"] = df["target"].astype(int)
except ValueError:
logger.warning("target not converted to int")
df.sort_values(by="target", inplace=True)
df["target"] = df["target"].astype(str)
# Radio - Display top features
if radio == "top":
top_features = df[fi["index"][0:4].values]
top_features["target"] = df["target"]
if len(top_numericals):
px_mat = px.scatter_matrix(top_features, color="target", height=800)
# C = ['rgb(166,206,227)', 'rgb(31,120,180)', 'rgb(178,223,138)',
# 'rgb(51,160,44)', 'rgb(251,154,153)', 'rgb(227,26,28)']
# N = len(df['target'].unique())
# matrix = ff.create_scatterplotmatrix(top_features, diag='box',
# index='target',
# title="",
# #colormap=C,
# colormap_type=cmap_type,
#
# height=800, width=900)
px_mat.update_traces(diagonal_visible=False)
graph = dcc.Graph(figure=px_mat)
else:
d = top_features
parcats = [
go.Parcats(
dimensions=[
{"label": column, "values": list(d[column].values)}
for column in d.columns
],
line={"color": y, "colorscale": "Portland"},
hoveron="color",
hoverinfo="count+probability",
arrangement="freeform",
)
]
layout = go.Layout(autosize=False, height=800)
fig = go.Figure(data=parcats, layout=layout)
graph = dcc.Graph(figure=fig)
elif radio == "numeric": # Top numeric features
if len(top_numericals):
df_num = df[top_numericals]
df_num["target"] = df["target"]
px_mat = px.scatter_matrix(df_num, color="target", height=800)
# matrix = ff.create_scatterplotmatrix(df_num, diag='box', #'box'
# index='target',
# title="",
# #colormap=C,
# colormap_type=cmap_type,
# height=1000, width=900)
graph = dcc.Graph(figure=px_mat)
px_mat.update_traces(diagonal_visible=False)
else:
graph = html.P("No numericals found")
elif radio == "nominal":
if len(top_nominals):
df_nom = df[top_nominals]
df_nom["target"] = df["target"]
parcats = [
go.Parcats(
dimensions=[
{"label": column, "values": list(df_nom[column].values)}
for column in df_nom.columns
],
line={
"color": pd.Categorical(df_nom["target"]).codes,
"colorscale": "Portland",
},
hoveron="color",
hoverinfo="count+probability",
arrangement="freeform",
)
]
layout = go.Layout(autosize=False, height=800)
fig = go.Figure(data=parcats, layout=layout)
graph = dcc.Graph(figure=fig)
else:
graph = html.P("No nominals found")
return html.Div(graph, className="twelve columns")
def feature_interactions(rows, radio, url, dummy):
data_id = int(re.search('data/(\d+)', url).group(1))
if dummy == "done":
df = pd.read_pickle('cache/df' + str(data_id) + '.pkl')
fi = pd.read_pickle('cache/fi' + str(data_id) + '.pkl')
else:
return []
meta_data = pd.DataFrame(rows)
try:
target_attribute = meta_data[meta_data["Target"] ==
"true"]["Attribute"].values[0]
target_type = (
meta_data[meta_data["Target"] == "true"]["DataType"].values[0])
except IndexError:
return "No target found", "No target found"
if target_type == "nominal" or target_type == "string":
y = pd.Categorical(df[target_attribute]).codes
else:
y = df[target_attribute]
# Feature interaction plots
df = clean_dataset(df)
numerical_features = list(
meta_data["Attribute"][meta_data["DataType"] == "numeric"])
nominal_features = list(
meta_data["Attribute"][meta_data["DataType"] == "nominal"])
top_numericals = (
fi['index'][fi['index'].isin(numerical_features)][:5])
top_nominals = (fi['index'][fi['index'].isin(nominal_features)][:5])
df['target'] = df[target_attribute]
C = [
'rgb(166,206,227)', 'rgb(31,120,180)', 'rgb(178,223,138)',
'rgb(51,160,44)', 'rgb(251,154,153)', 'rgb(227,26,28)'
]
if target_type == "numeric":
cmap_type = 'seq'
df['target'] = y
df['target'] = pd.cut(df['target'], 1000).astype(str)
cat = df['target'].str.extract('\((.*),',
expand=False).astype(float)
df['bin'] = pd.Series(cat)
df.sort_values(by='bin', inplace=True)
df.drop('bin', axis=1, inplace=True)
else:
cmap_type = 'cat'
N = len(df['target'].unique())
try:
df['target'] = df['target'].astype(int)
except ValueError:
print("target not converted to int")
df.sort_values(by='target', inplace=True)
df['target'] = df['target'].astype(str)
if radio == "top":
top_features = df[fi['index'][0:5].values]
top_features['target'] = df['target']
if len(top_numericals):
matrix = ff.create_scatterplotmatrix(
top_features,
title='Top feature interactions',
diag='box',
index='target',
#colormap=C,
colormap_type=cmap_type,
height=800,
width=900)
graph = dcc.Graph(figure=matrix)
else:
d = top_features
parcats = [
go.Parcats(dimensions=[{
'label': column,
'values': list(d[column].values)
} for column in d.columns],
line={
'color': y,
'colorscale': 'Portland'
},
hoveron='color',
hoverinfo='count+probability',
arrangement='freeform')
]
layout = go.Layout(autosize=False, width=1200, height=800)
fig = go.Figure(data=parcats, layout=layout)
graph = dcc.Graph(figure=fig)
elif radio == "numeric":
if len(top_numericals):
df_num = df[top_numericals]
df_num['target'] = df['target']
matrix = ff.create_scatterplotmatrix(
df_num,
title='Top numeric feature interactions',
diag='box',
index='target',
#colormap=C,
colormap_type=cmap_type,
height=1000,
width=1000)
graph = dcc.Graph(figure=matrix)
else:
graph = html.P("No numericals found")
elif radio == "nominal":
if len(top_nominals):
df_nom = df[top_nominals]
df_nom['target'] = df['target']
parcats = [
go.Parcats(dimensions=[{
'label':
column,
'values':
list(df_nom[column].values)
} for column in df_nom.columns],
line={
'color':
pd.Categorical(df_nom['target']).codes,
'colorscale': 'Portland'
},
hoveron='color',
hoverinfo='count+probability',
arrangement='freeform')
]
layout = go.Layout(autosize=False, width=1000, height=800)
fig = go.Figure(data=parcats, layout=layout)
graph = dcc.Graph(figure=fig)
else:
graph = html.P("No nominals found")
return html.Div(graph)
def g_aluvial_cat(param_data, param_theme, param_dims):
"""
Parameters
----------
param_data : pd.DataFrame : data frame con tabla a graficar (tabla 3)
param_theme : dict : diccionario con tema de visualizaciones
param_dims : dict : diccionario con tamanos para visualizaciones
Returns
-------
fig_g_aluvial_cat : plotly : objeto/diccionario tipo plotly para graficar
Debugging
---------
param_data = tabla_3
param_theme = tema_base
param_dims = dimensiones_base
"""
# generacion de dimension: categoria
categoria_dim = go.parcats.Dimension(
values=param_data['categoria'],
label='categoria')
# generacion de dimension: pais
pais_dim = go.parcats.Dimension(
values=param_data['pais'],
label='pais')
# generacion de dimension: frecuencia de ocurrencia
frecuencia_dim = go.parcats.Dimension(
values=param_data['frecuencia'],
label='frecuencia')
# generacion de dimension: presencia de patrones tipo 1
tipo_1_dim = go.parcats.Dimension(
values=param_data['tipo_1'],
label="tipo_1",
categoryarray=[0, 1],
ticktext=['sin patron', 'con patron'])
# generacion de dimension: presencia de patrones tipo 2
tipo_2_dim = go.parcats.Dimension(
values=param_data['tipo_2'],
label="tipo_2",
categoryarray=[0, 1],
ticktext=['sin patron', 'con patron'])
# generacion de dimension: presencia de patrones tipo 3
tipo_3_dim = go.parcats.Dimension(
values=param_data['tipo_3'],
label="tipo_3",
categoryarray=[0, 1],
ticktext=['sin patron', 'con patron'])
# vector de colores para todas las lineas
colores = [param_theme['color_linea_9'], param_theme['color_linea_2'],
param_theme['color_linea_3'], param_theme['color_linea_4'],
param_theme['color_linea_5'], param_theme['color_linea_6'],
param_theme['color_linea_7'], param_theme['color_linea_8'],
param_theme['color_linea_1']]
# crear columna de color en los datos de entrada
param_data['color'] = ['#ABABAB']*len(param_data['id'])
for i in range(0, len(param_data['categoria'])):
if param_data['categoria'].iloc[i] == 'Tasas de interes':
param_data['color'].iloc[i] = colores[0]
elif param_data['categoria'].iloc[i] == 'actividad economica':
param_data['color'].iloc[i] = colores[3]
elif param_data['categoria'].iloc[i] == 'consumo':
param_data['color'].iloc[i] = colores[8]
elif param_data['categoria'].iloc[i] == 'energia':
param_data['color'].iloc[i] = colores[6]
elif param_data['categoria'].iloc[i] == 'flujos de capital':
param_data['color'].iloc[i] = colores[4]
elif param_data['categoria'].iloc[i] == 'inflacion':
param_data['color'].iloc[i] = colores[5]
elif param_data['categoria'].iloc[i] == 'mercado inmobiliario':
param_data['color'].iloc[i] = colores[1]
elif param_data['categoria'].iloc[i] == 'mercado laboral':
param_data['color'].iloc[i] = colores[7]
elif param_data['categoria'].iloc[i] == 'subasta de bonos':
param_data['color'].iloc[i] = colores[2]
color = param_data['color'].tolist()
# generacion del objeto figura
fig_g_aluvial_cat = go.Figure()
# agregar trazo de grafica tipo aluvial (parallel categories)
fig_g_aluvial_cat.add_trace(go.Parcats(
dimensions=[categoria_dim, frecuencia_dim, pais_dim,
tipo_1_dim, tipo_2_dim, tipo_3_dim],
line={'color': color},
hoveron='color', hoverinfo='count+probability',
labelfont={'size': 14, 'family': 'Times',
'color': param_theme['color_texto_ejes']},
tickfont={'size': 14, 'family': 'Times',
'color': param_theme['color_texto_ejes']},
arrangement='perpendicular'))
# layout de margen, titulos y ejes
fig_g_aluvial_cat.update_layout(
margin=go.layout.Margin(l=100, r=25, b=5, t=25, pad=10),
title=None)
# Formato de tamanos
fig_g_aluvial_cat.layout.autosize = True
fig_g_aluvial_cat.layout.width = param_dims['figura_3']['width']
fig_g_aluvial_cat.layout.height = param_dims['figura_3']['height']
return fig_g_aluvial_cat
dims = []
for _, dim_key in enumerate(dim_parallel):
dims.append(
go.parcats.Dimension(values=filtered_table[dim_key],
label=dim_key))
if c_key != 'None':
unique_list = np.sort(filtered_table[c_key].unique())
if np.issubdtype(unique_list.dtype, np.integer) or \
np.issubdtype(unique_list.dtype, np.floating):
parallel_fig = go.Figure(data=[
go.Parcats(dimensions=dims,
line={
'color': filtered_table[c_key],
'colorbar': dict(title=c_key)
},
hoveron='color',
hoverinfo='count+probability',
arrangement='freeform')
])
else:
filtered_table['_C_'] = np.zeros_like(
filtered_table[c_key])
for idx, var in enumerate(unique_list):
filtered_table.loc[filtered_table[c_key] == var,
'_C_'] = idx
parallel_fig = go.Figure(data=[
go.Parcats(dimensions=dims,
line={'color': filtered_table['_C_']},
hoverinfo='count+probability',
