def check_normal_dist(x: pd.Series):
from scipy import stats
from scipy.stats import norm
if x.dtype == 'object':
return
fig, ax = _visu._create_fig(ncols=2)
sns.distplot(x, fit=norm, ax=ax[0])
stats.probplot(x, plot=ax[1])
_visu._plot()
def value_count(data: pd.Series):
"""
Plot value counts
"""
fig, ax = _visu._create_fig()
counts = data.value_counts()
ax.bar(counts.index, counts.values)
ax.set(title=f'Value Count: {counts.name}')
_visu._plot()
def histgram(data: pd.Series, bins: Optional[int] = None):
if data.dtype == 'object':
print("Ignore non number type")
return
bin_number = bins
if bin_number is None:
bin_number = int(1 + np.log2(len(data)))
fig, ax = _visu._create_fig()
ax.hist(data, bins=bin_number)
ax.set(title=f'Histgram {data.name}(bin:{bin_number})',
xlabel=data.name,
ylabel='Frequency')
_visu._plot()
def density_plot(data: pd.Series, band_width: Optional[float] = None):
band = band_width
if data.dtype == 'object':
print("Ignore non number type")
return
if band is None:
band = 'scott'
fig, ax = _visu._create_fig()
sns.kdeplot(data, shade=True, ax=ax, bw=band)
ax.set(title=f'Density {data.name}(band width:{band})',
xlabel=data.name,
ylabel='Frequency')
_visu._plot()
def density_2d_plot(x: pd.Series,
y: pd.Series,
band_width: Optional[float] = None):
band = band_width
if x.dtype == 'object' or y.dtype == 'object':
print("Ignore non number type")
return
if band is None:
band = 'scott'
fig, ax = _visu._create_fig()
sns.kdeplot(x, y, shade=True, ax=ax, bw=band)
ax.set(title=f'Density {x.name} x {y.name} (band width:{band})',
xlabel=x.name,
ylabel=y.name)
_visu._plot()
def pareto(data: pd.Series):
"""
Plot pareto chart
"""
if data.dtype != 'object':
return
fig, ax = _visu._create_fig()
counts = data.value_counts()
ax.bar(counts.index, counts.values, label='count')
ax.set(title=f'Pareto: {counts.name}')
ax2 = ax.twinx()
ratio = counts / counts.sum()
ax2.plot(ratio.index, ratio.cumsum(), label='ratio', c=palette[1])
ax2.set(ylim=(0, 1.05))
h1, l1 = ax.get_legend_handles_labels()
h2, l2 = ax2.get_legend_handles_labels()
_visu.plt.legend(h1 + h2, l1 + l2)
_visu._plot()
def roccurve(y, y_prob):
"""
plot ROC
"""
from sklearn.metrics import roc_auc_score
from sklearn.metrics import roc_curve
auc = roc_auc_score(y, y_prob)
fpr, tpr, thresholds = roc_curve(y, y_prob)
fig, ax = _visu._create_fig()
ax.plot([0, 1e-10, 1], [0, 1, 1], color=palette[0], label='Perfect')
ax.plot(fpr, tpr, color=palette[1], label='Model')
ax.plot([0, 1], [0, 1],
linestyle='dashed', color=palette[2], label='Random')
ax_params = {
"title": f"ROC curve (AUC:{auc:.04})",
"xlabel": "FP rate",
"ylabel": "TP rate",
}
ax.set(**ax_params)
ax.legend()
_visu._plot()
pass
def capcurve(y, y_prob):
"""
plot CAP
"""
from scipy import integrate
num_sum = np.sum(y)
num_count = len(y)
rate_val = float(num_sum) / float(num_count)
ideal = np.array([[0, rate_val, 1], [0, 1, 1]])
y_cap_df = (pd.DataFrame({'y': y, 'yp': y_prob})
.sort_values('yp', ascending=False)
.reset_index(drop=True))
y_cap_df['y_rate'] = np.cumsum(y_cap_df.y)/num_sum
y_cap_df['x_rate'] = np.arange(num_count)/num_count
perfect = integrate.simps(ideal[1, :], ideal[0, :])
model = integrate.simps(y_cap_df.y_rate, y_cap_df.x_rate)
_random = integrate.simps(y_cap_df.x_rate, y_cap_df.x_rate)
gini = (model - _random) / (perfect - _random)
fig, ax = _visu._create_fig()
ax.plot(ideal[0, :], ideal[1, :],
color=palette[0], label='Perfect')
ax.plot(y_cap_df.x_rate, y_cap_df.y_rate,
color=palette[1], label='Model')
ax.plot(y_cap_df.x_rate, y_cap_df.x_rate,
linestyle='dashed', color=palette[2], label='Random')
ax_params = {
"title": f"CAP Curve (Gini index:{gini:.04})",
"xlabel": "data ratio",
"ylabel": "positive ratio",
}
ax.set(**ax_params)
ax.legend()
_visu._plot()
def box_plot(x: pd.Series, y: pd.Series):
fig, ax = _visu._create_fig()
sns.boxplot(x=x, y=y)
sns.stripplot(x=x, y=y, color='black', size=2, jitter=1, alpha=0.5)
ax.set(title=f'Box {x.name} x {y.name}', xlabel=x.name, ylabel=y.name)
_visu._plot()
def violin_plot(x: pd.Series, y: pd.Series):
fig, ax = _visu._create_fig()
sns.violinplot(x=x, y=y, scale='width', inner='quartile')
ax.set(title=f'Violin {x.name} x {y.name}', xlabel=x.name, ylabel=y.name)
_visu._plot()
def scatter_plot(x: pd.Series, y: pd.Series):
fig, ax = _visu._create_fig()
ax.scatter(x, y)
ax.set(title=f'Scatter {x.name} x {y.name}', xlabel=x.name, ylabel=y.name)
_visu._plot()