def _box_plot(df, name, target, category, section_number, sub_section, display_name):
display_target = get_display_text(target)
display_category = get_display_text(category)
sub_section = util.header(section_number, sub_section,
"{} Box Plot With {} by {}".format(display_name, display_target, display_category))
try:
dskc_graphs.box_plot(df, category, target, name, flip=True, grouped=True, agg_func="mean")
except:
plt.show()
print("\nNot available.\n")
return sub_section
def boolean_col(df, name, target, target_true=False, section_number=1):
# get names
display_name = get_display_text(name)
sub_section = 1
# set series
series = df[name]
# bars
sub_section = _bars(series, section_number, sub_section, display_name)
# bars proportion
if target:
display_target = get_display_text(target)
_bars_proportion(df, name, target, target_true, section_number,
sub_section, display_name, display_target)
def date_col(df, name, target, target_true=False, section_number=1):
# set names
display_name = get_display_text(name)
sub_section = 1
# set series
series = df[name]
# graphs
if str(series.dtype).find("date") >= 0:
return
markdown_h2("Line graph")
# todo line
if name.lower().endswith("_weekday"):
xticks = get_weekdays()
else:
xticks = False
# bars
sub_section = _bars(series, xticks, section_number, sub_section,
display_name)
# bars proportion
if target:
_bars_proportion(df, name, xticks, target, target_true, section_number,
sub_section, display_name)
def number_col(df, name, target, section_number=1, categories=[]):
'''
:param df: pandas dataframe
:param name: name of the column
:param target: target column variable
:return:
'''
# set names
display_name = get_display_text(name)
sub_section = 1
# set series
series = df[name]
# correlation
sub_section = _correlation(df, name, section_number, sub_section, display_name)
# histogram
sub_section = _hist(series, section_number, sub_section, display_name)
# density plot
if target:
sub_section = _dns_plot_target(df, name, target, section_number, sub_section, display_name)
else:
sub_section = _dns_plot(df, name, section_number, sub_section, display_name)
# histogram no outliers
sub_section = _hist_no_outliers(series, section_number, sub_section, display_name)
# desnity plot no ouliers
if target:
_dns_plot_target_no_outliers(df, name, target, section_number, sub_section, display_name)
else:
_dns_plot_no_outliers(df, name, section_number, sub_section, display_name)
# box plots
if not target:
return
for category in categories:
if category == target:
continue
sub_section = _box_plot(df, name, target, category, section_number, sub_section, display_name)
def text_col(df,
name,
target=None,
target_true=False,
section_number=1,
top_words=15,
stop_words=[]):
# get names
display_name = get_display_text(name)
sub_section = 1
# set series
series = df[name]
# wordcloud
sub_section = _wordcloud(series, section_number, sub_section, display_name,
stop_words)
# bars graph
text = get_text_from(series, stop_words=stop_words)
# set word series
words = text.split(" ")
words_series = pd.Series(words)
# top n words
sub_section = _top_words(words_series, top_words, section_number,
sub_section, display_name)
# text proportion graphs
if not target is None:
try:
_text_proportion_succcess(series, words_series, df[target],
target_true, top_words, section_number,
sub_section, display_name)
except:
plt.show()
print("\nNot available.\n")
def box_plot(df,
x_col,
fill_col,
agg_col,
agg_func='count',
filter_col=None,
filterby=[""],
no_outliers=False,
title=None,
ylabel=None,
grouped=True,
flip=True,
dots=True):
'''
:param df: dataframe,x_col: variable located on x axis,fill_col: variable for coloring the dots,agg_col: name of the column to which the aggregation function is going to perform
:optional agg_fun: for example: sum,min,mean,median,max,etc. filter_col: after groupying filtering option,filterby: list of characters to be filtered out (located in filter column),outliers,title
:return: ggplot graphs of univariate analysis
:type: boxplot
:usage: after cleaning dataframe
'''
# set display texts
fill_label = get_display_text(fill_col)
if grouped:
aux, aux_str = _get_grouped_aux(df, x_col, fill_col, agg_col, agg_func,
filter_col, filterby)
else:
aux = df
aux_str = agg_col
# sort x labels
aux = aux.sort_values(by=[x_col])
xcol_list = aux.astype(str)[x_col].unique().tolist()
# graph creation
graph = (ggplot(aux) + geom_boxplot(aes(x=x_col, y=aux_str)) + theme_bw() +
theme(axis_line_x=element_line(color='gray'),
axis_line_y=element_line(color='gray'),
line=element_line(color='white')) +
scale_fill_manual(values=colors.OFICIAL_COLORS, name=fill_label) +
scale_x_discrete(limits=xcol_list) +
scale_y_continuous(labels=custom_format('{:,.0f}')))
# dots
if dots:
graph += geom_jitter(aes(x=x_col, y=aux_str, fill=fill_col))
# no outliers
aux_describe = aux[aux_str].describe().loc
if no_outliers:
graph += ylim(aux_describe["min"], aux_describe["75%"])
# title
if title != None:
graph += ggtitle("Box plot:" + str(title))
# flip
if flip:
graph += coord_flip()
# set y label
if not ylabel:
ylabel = get_display_text(agg_col)
graph += ylab(ylabel)
# show
graph.draw()
plt.show()