美赛用到的可视化技巧

前言

这篇就是纯粹给自己记录的，写的非常乱，建议是别看

以下的图都得加上一条语句

pyplot.style.use('seaborn')

敏感度分析

import numpy as np
import pandas as pd
from pyecharts import options as opts
from pyecharts.charts import Bar3D
from pyecharts.faker import Faker
import random
from pyecharts.globals import ThemeType
 
data = [(i, j, j*0.2+3+np.random.randint(2)) for i in range(21) for j in range(21)]
bar_3d=(
    Bar3D(init_opts=opts.InitOpts(width="1000px", height="1000px"))
        .add('sensitivity',
            data=[[d[0],d[1],d[2]] for d in data],#传入数据的索引
            xaxis3d_opts=opts.Axis3DOpts(type_='category', name='GOLD', interval=2),#传入x轴的数据以及类型
            yaxis3d_opts=opts.Axis3DOpts(type_='category', name='BITCOIN', interval=2),#传入y轴的数据以及类型
            zaxis3d_opts=opts.Axis3DOpts(type_='value', name='VALUE')).set_global_opts(visualmap_opts=opts.VisualMapOpts(max_=25)
        )
        .set_global_opts(
            title_opts=opts.TitleOpts("标准3D柱状图"),
            visualmap_opts=opts.VisualMapOpts(
                max_=9,
                range_color=[
                    "#313695",
                    "#4575b4",
                    "#74add1",
                    "#abd9e9",
                    "#e0f3f8",
                    "#ffffbf",
                    "#fee090",
                    "#fdae61",
                    "#f46d43",
                    "#d73027",
                    "#a50026",
                ],
            )
        ).render('灵敏度分析.html')
)

Xgboost画特征重要性

import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
from xgboost import XGBRegressor, plot_importance
from sklearn.feature_selection import SelectFromModel
plt.style.use('seaborn')

X_gold = gold_VR.iloc[:, 2:]
y_gold = gold_VR.iloc[:,1]
# fit model no training data
model_gold = XGBRegressor()
model_gold.fit(X_gold, y_gold)
# plot feature importance
_, ax = plt.subplots(1,1,figsize=[12, 8])
plot_importance(model_gold, ax=ax, importance_type='gain')
plt.show()

print('特征重要性数值 : ', model_gold.feature_importances_)
fig = plt.figure(figsize=[12, 8])
col_names = list(X_gold.columns)
ax.xaxis.set_major_formatter(ticker.FuncFormatter(lambda x : col_names[int(x)]))
plt.setp(plt.gca().get_xticklabels(), rotation=45, horizontalalignment='right')
plt.bar(col_names, model_gold.feature_importances_)
plt.show()

设置坐标轴格式

# 日期
date_range = pd.read_csv('../data/gold_complete.csv', index_col=0)[-800:]['Date'].values
# 可视化
fig = plt.figure(figsize=[12, 8])
ax = fig.add_subplot(111)
for i in range(data.shape[1]):
    ax.plot(np.arange(1, data.shape[0]+1), data.iloc[:,i], label = data.columns[i])

# 设置坐标轴格式
def format_date(x, pos=None):
    if x < 0 or x > data.shape[0]:
        return ''
    elif x==800:
        return date_range[-1]
    return date_range[int(x)]
ax.xaxis.set_major_formatter(ticker.FuncFormatter(format_date))
plt.legend()
plt.setp(plt.gca().get_xticklabels(), rotation=45, horizontalalignment='right')
plt.show()

堆叠柱状图

# 日期
date_range = pd.read_csv('../data/gold_complete.csv', index_col=0)[-50:]['Date'].values
# 可视化
fig = plt.figure(figsize=[12, 8])
ax = fig.add_subplot(111)

lens = 50
barWidth = 0.85
r = range(len(Dollar))
ax.bar(r[-lens:], Dollar[-lens:], width=barWidth, label = 'Dollr',color = '#ffce2b')
ax.bar(r[-lens:], Gold[-lens:], bottom=Dollar[-lens:], width=barWidth, label = 'Gold',color='#00c5d2')
ax.bar(r[-lens:], BitCoin[-lens:], 
    bottom=[i+j for i,j in zip(Dollar[-lens:], Gold[-lens:])], width=barWidth, label = 'BitCoin',color='#ff812c')

# 设置坐标轴格式
def format_date(x, pos=None):
    x = x +1025
    if x < 0 or x > len(date_range):
        return ''
    return date_range[int(x)]
ax.xaxis.set_major_formatter(ticker.FuncFormatter(format_date))
plt.legend(loc = 'best')
plt.setp(plt.gca().get_xticklabels(), rotation=45, horizontalalignment='right')
plt.ylabel('Ratio')
plt.show()

绘制带有置信度的趋势图

def fit_lstm(data_prepare, n_neurons=50, n_batch=72, n_epoch=100, loss='mae', optimizer='adam', repeats=1):
    train_X = data_prepare[2]
    train_y = data_prepare[3]
    test_X = data_prepare[4]
    test_y = data_prepare[5]
    model_list = []
    loss_list = []
    val_loss_list = []
    for i in range(repeats):
        print('training mode : ', i)
        #设计神经网络
        model = Sequential()
        model.add(LSTM(n_neurons, input_shape=(train_X.shape[1], train_X.shape[2])))
        model.add(Dense(train_y.shape[1]))
        model.compile(loss=loss, optimizer=optimizer)
        #拟合神经网络
        history = model.fit(train_X, train_y, epochs=n_epoch, batch_size=n_batch, validation_data=(test_X, test_y), verbose=0, shuffle=False)
        #画出学习过程
        loss_list.append(history.history['loss'])
        val_loss_list.append(history.history['val_loss'])
        #保存model
        model_list.append(model)
    train_loss = np.array(loss_list)
    val_loss = np.array(val_loss_list)
    train_mean = np.mean(train_loss, axis=0)
    train_std = np.std(train_loss, axis=0)
    val_mean = np.mean(val_loss, axis=0)
    val_std = np.std(val_loss, axis=0)
    x_list = range(len(loss_list[0]))

    plt.figure(figsize=[12,8])
    plt.plot(x_list, train_mean, color = 'blue', label='train')
    plt.fill_between(x_list, train_mean+train_std, train_mean-train_std, alpha=0.15,color='blue')
    plt.plot(x_list, val_mean, color = 'red', label='train')
    plt.fill_between(x_list, val_mean+val_std, val_mean-val_std, alpha=0.15,color='red')
    plt.legend(["train","test"])
    plt.xlabel('epoch')
    plt.ylabel('loss value')
    plt.show()
    return model_list

可视化缺失值

import missingno as mno  
mno.matrix(gold_data)