def barchart(dataframe, numerical_column, is_percent=False):
"""
df: frequency table
"""
rcParams['figure.figsize'] = 8, 5
if (is_percent):
## removing the % sign
# dataframe[numerical_column] = dataframe[numerical_column].str.rstrip('%').astype('float')
# sns.barplot(x=numerical_column, y=dataframe.index, data=dataframe, orient="h", order=dataframe.index, color="#337ab7")
dataframe.sort_values(numerical_column, ascending=True)[numerical_column].plot.barh(color="#337ab7")
# plt.barh(dataframe.index.values, dataframe[numerical_column])
plt.xlim(0, 100)
else:
# sns.barplot(x=numerical_column, y=dataframe.index, data=dataframe, orient="h", order=dataframe.index, color="#337ab7")
dataframe.sort_values(numerical_column, ascending=True)[numerical_column].plot.barh(color="#337ab7")
# df.plot.barh()
# plt.subplots_adjust(left=0.1, right=0.9, top=0.7, bottom=0.2)
plt.xticks(fontsize=12, rotation=0)
plt.yticks(fontsize=12, rotation=0)
plt.xlabel(numerical_column, fontsize=15)
plt.ylabel(dataframe.index.name, fontsize=15)
# plt.subplots_adjust(left=0.4, right=0.6, top=0.9, bottom=0.1)
plt.tight_layout(rect=(0.1, 0.1, 0.9, 0.9))
return plot_360_n0sc0pe(plt)
def stackbarchart(df, cat_a, cat_b):
rcParams['figure.figsize'] = (10, 4)
## keeping only top 10 categories
temp_df = df[[cat_a, cat_b]]
sorted_order = list(temp_df[cat_a].value_counts().index)
if (df[cat_a].nunique() > 10):
temp_df.loc[~temp_df[cat_a].isin(sorted_order[0:9]), cat_a] = "Others"
sorted_order = sorted_order[0:9] + ["Others"]
if (df[cat_b].nunique() > 10):
top_categories = list(temp_df[cat_b].value_counts().index[0:9])
temp_df.loc[~temp_df[cat_b].isin(top_categories), cat_b] = "Others"
## plots
fig, (ax1, ax2) = plt.subplots(1, 2)
grouped_df = temp_df.groupby([cat_a, cat_b]).size()
grouped_df.unstack().reindex(index=sorted_order).plot(kind='bar', stacked=True, ax = ax1, legend=False)
ax1.set_title("Bar Chart",weight="bold").set_fontsize('12')
cross_df = pd.crosstab(temp_df[cat_a], temp_df[cat_b]).apply(lambda r: r/r.sum(), axis=1)
cross_df.reindex(index=sorted_order).plot(kind='bar', stacked=True, ax=ax2 )
ax2.set_title("Stacked Chart",weight="bold").set_fontsize('12')
ax2.legend(title=cat_b, loc='center left', bbox_to_anchor=(1, 0.5))
plt.tight_layout(rect=(0.1, 0.05, 0.9, 0.95))
return plot_360_n0sc0pe(plt)
def missing_count_row_wise(df):
rcParams['figure.figsize'] = 10, 4
dataframe = df
# def missing_count_row_wise(dataframe)
dataframe = df
column_count = len(df.columns)
row_count = dataframe.shape[0]
## missing: column wise
missing_row_df = pd.DataFrame(
dataframe.isnull().sum(axis=1).value_counts(),
columns=["Number of rows"]).reset_index()
missing_row_df["percent"] = missing_row_df["index"].apply(
lambda x: 100 * round(x / column_count, 2)).astype(int)
missing_row_df["Missing cells %"] = missing_row_df.apply(
lambda x: str(x["index"]) + " (" + str(x["percent"]) + "%)", axis=1)
missing_row_df.sort_values("index", ascending=False, inplace=True)
sns.barplot(x="Number of rows",
y="Missing cells %",
data=missing_row_df,
orient="h",
color="#337ab7")
plt.xlabel('Number of rows', fontsize='12')
plt.ylabel('Number of missing cells', fontsize='12')
# plt.title("Missing values % in each column",weight="bold", fontsize='15')
plt.tight_layout(rect=(0.1, 0, 0.9, 1))
return plot_360_n0sc0pe(plt)
def scatterplot(df, num_a, num_b):
## plots
fig, (ax1, ax2) = plt.subplots(1, 2)
sns.distplot( df[num_a].dropna() , color="skyblue", label=num_a, ax=ax1)
sns.distplot( df[num_b].dropna() , color="red", label=num_b, ax=ax1)
ax1.set_title("",weight="bold").set_fontsize('12')
sns.regplot(df[num_a], df[num_b], ax=ax2)
reg_coeff = round(df[num_a].corr(df[num_b]), 2)
ax2.set_title("R: " + str(reg_coeff), weight="bold").set_fontsize('12')
return plot_360_n0sc0pe(plt)
def wordcloud(series):
stopwords = set(STOPWORDS)
wordcloud = WordCloud(
background_color='white',
stopwords=stopwords,
max_words=200,
max_font_size=40,
random_state=42
).generate(str(series.values))
fig = plt.figure(1)
plt.imshow(wordcloud)
plt.axis('off')
return plot_360_n0sc0pe(plt)
def missing_matrix(data: pd.DataFrame) -> str:
"""Generate missing values matrix plot
Args:
data: Pandas DataFrame to generate missing values matrix from.
Returns:
The resulting missing values matrix encoded as a string.
"""
labels = True
missingno.matrix(
data,
figsize=(10, 4),
color=hex_to_rgb("#337ab7"),
fontsize=10,
sparkline=False,
labels=labels,
)
plt.subplots_adjust(left=0.1, right=0.9, top=0.7, bottom=0.2)
return plot_360_n0sc0pe(plt)
def missing_count_column_wise(df):
rcParams['figure.figsize'] = 10, 4
dataframe = df
row_count = dataframe.shape[0]
## missing: column wise
missing_column_df = dataframe.isnull().sum()
missing_column_df = pd.DataFrame(missing_column_df,
columns=["Missing values"])
missing_column_df["values"] = row_count
missing_column_df = missing_column_df.apply(
lambda x: 100 * round(x / row_count, 2)).astype(int)
missing_column_df.sort_values("Missing values",
ascending=False,
inplace=True)
## bar plot
sns.set_style("darkgrid")
## plot 1 - "total" - (top) series
sns.barplot(x=missing_column_df.index,
y="values",
data=missing_column_df,
color="#337ab7",
label="Values")
## plot 2 - overlay - "bottom" series
sns.barplot(x=missing_column_df.index,
y="Missing values",
data=missing_column_df,
color="red",
label="Missing")
plt.legend(loc='center left', bbox_to_anchor=(1, 0.5))
plt.xticks(rotation="90")
plt.ylabel('(%)', fontsize='12')
plt.title("Missing values % in each column",
weight="bold").set_fontsize('12')
plt.tight_layout()
return plot_360_n0sc0pe(plt)
def boxplot(df, num, cat):
rcParams['figure.figsize'] = 10, 5
temp_df = df[[num, cat]]
sorted_order = temp_df.groupby(cat, as_index=False)[num].mean().sort_values(num, ascending=False)
if (sorted_order.shape[0] > 10):
top_categories = list(sorted_order[cat][0:9])
temp_df.loc[~temp_df[cat].isin(top_categories), cat] = "Others"
sns.boxplot(x=cat, y=num, data=temp_df, order = top_categories + ["Others"])
else:
sns.boxplot(x=cat, y=num, data=temp_df, order = sorted_order[cat])
plt.xticks(fontsize=12, rotation=0)
plt.yticks(fontsize=12, rotation=0)
plt.xlabel(cat, fontsize=15)
plt.ylabel(num, fontsize=15)
# plt.subplots_adjust(left=0.4, right=0.6, top=0.9, bottom=0.1)
plt.tight_layout(rect=(0.1, 0.1, 0.9, 0.9))
return plot_360_n0sc0pe(plt)
def histogram(series: pd.Series, col_name):
## drawing
## - fixing the size of the figure
rcParams['figure.figsize'] = 10, 5
x = np.array(series.dropna())
# Cut the window in 2 parts
f, (ax_box, ax_hist) = plt.subplots(2, sharex=True,
gridspec_kw={"height_ratios": (.15, .85)}
)
sns.boxplot(x, ax=ax_box, color="#337ab7")
sns.distplot(x, ax=ax_hist, color="#337ab7")
ax_box.set(yticks=[])
sns.despine(ax=ax_hist)
sns.despine(ax=ax_box, left=True)
plt.xticks(fontsize=12, rotation=0)
plt.yticks(fontsize=12, rotation=0)
plt.tight_layout()
return plot_360_n0sc0pe(plt)
def networkplot(column_types, associations):
## preprocessing assns
assn_df = associations.copy()
assn_df["assn"] = abs(assn_df["association"])
## - filter
assn_df = assn_df[assn_df["assn"] >= .01]
## creating graph
G = nx.from_pandas_edgelist(assn_df,
source='col_a',
target='col_b',
edge_attr=["assn", "type_"],
create_using=nx.DiGraph())
## edges
corr_edges = [(u, v) for (u, v, d) in G.edges(data=True)
if d['type_'] == "NUM-NUM"]
corr_ratio_edges = [(u, v) for (u, v, d) in G.edges(data=True)
if d['type_'] in ("NUM-CAT", "CAT-NUM")]
theil_edges = [(u, v) for (u, v, d) in G.edges(data=True)
if d['type_'] in ("CAT-CAT")]
## size and color attribute of nodeS
size_df = assn_df.groupby(
"col_a", as_index=False)["assn"].sum().rename(columns={"assn": "size"})
## - normalizing
max_size = size_df["size"].max()
size_df["size"] = size_df["size"] / max_size
for i in list(G.nodes()):
G.nodes[i]['size'] = size_df[size_df['col_a'] == i]['size'].values[0]
if (column_types[i] == "BOOL"):
G.nodes[i]['color'] = "#c03d3e"
elif (column_types[i] == "CAT"):
G.nodes[i]['color'] = "#3a923a"
elif (column_types[i] == "NUM"):
G.nodes[i]['color'] = "#337ab7"
else:
G.nodes[i]['color'] = "blue"
## drawing
## - fixing the size of the figure
plt.figure(figsize=(10, 4))
## color, size, width
node_color = [nx.get_node_attributes(G, 'color')[v] for v in G]
node_size = [1000 * nx.get_node_attributes(G, 'size')[v] for v in G]
edge_width = [20 * G[u][v]['assn'] for u, v in G.edges()]
## layout
pos = nx.spring_layout(G, iterations=50)
# node labels
nx.draw_networkx_labels(G,
pos,
with_labels=True,
font_size=15,
font_family='sans-serif',
font_color="#000000",
font_weight="bold")
## nodes
nx.draw_networkx_nodes(G,
pos,
node_color=node_color,
node_size=node_size,
node_shape="o",
alpha=0.9,
linewidths=10)
## edges
# nx.draw_networkx_edges(G, pos, edgelist=corr_edges,
# width=edge_width, alpha=0.2, style='solid', edge_color="grey", arrows=False)
# nx.draw_networkx_edges(G, pos, edgelist=corr_ratio_edges,
# width=edge_width, alpha=0.2, style='solid', edge_color="grey", arrows=False)
# nx.draw_networkx_edges(G, pos, edgelist=theil_edges,
# width=edge_width, alpha=0.2, style='solid', edge_color="grey", arrows=False)
nx.draw_networkx_edges(G,
pos,
width=edge_width,
alpha=0.15,
style='solid',
edge_color="grey",
arrows=False)
## edge labels
# edge_labels =dict([((u, v), d['assn']) for u, v, d in G.edges(data=True)])
# nx.draw_networkx_edge_labels(G, pos, edge_labels=edge_labels)
plt.axis('off')
plt.tight_layout()
return plot_360_n0sc0pe(plt)