-
Notifications
You must be signed in to change notification settings - Fork 0
/
tsa.py
171 lines (162 loc) · 9.07 KB
/
tsa.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
# Local resources
import functions
# Data Preprocessing
import pandas as pd
import numpy as np
import json
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error
from sklearn.model_selection import TimeSeriesSplit
# Data Visualisation
import plotly.express as px
import plotly.graph_objs as go
# Available templates: ['ggplot2', 'seaborn', 'simple_white', 'plotly', 'plotly_white', 'plotly_dark', 'presentation',
# 'xgridoff', 'ygridoff', 'gridon', 'none']
px.defaults.template = "plotly_dark"
# Statistical tools for time series analysis
from statsmodels.tsa.stattools import adfuller, kpss, acf, pacf
from statsmodels.tsa.seasonal import seasonal_decompose
# Algorithms ML
from sklearn.linear_model import RidgeCV, ElasticNetCV, LassoCV
from sklearn.ensemble import RandomForestRegressor
from lightgbm import LGBMRegressor
from statsmodels.tsa.statespace.sarimax import SARIMAX
#------------------------------------------------- ПОДГОТОВКА ДАННЫХ---------------------------------------------------
data = pd.read_csv('./datasets/taxi.csv')
# приведение данных к DateTimeIndex
df = pd.read_csv('./datasets/taxi.csv', parse_dates=[0], index_col=[0])
# понижение частоты временного ряда
df.sort_index()
df_resample = df.resample('1H').sum()
fig_freq_1h = px.line(df_resample, x=df_resample.index, y=df_resample['num_orders'], title='Частота ряда 1 час')
# декомпозиция временного ряда
decomposed = seasonal_decompose(df_resample, model='additive')
trend = decomposed.trend
trend = trend.dropna()
seasonal = decomposed.seasonal
observed = decomposed.observed
resid = decomposed.resid
# переводим серию в датафрейм
df_trend = trend.to_frame(name='trend')
df_seasonal = seasonal.to_frame(name='seasonal')
df_resid = resid.to_frame(name='resid')
df_observed = observed.to_frame(name='observed')
# графики декомпозиции
fig_trend = px.line(trend, x=trend.index, y=trend)
fig_seasonal = px.line(seasonal, x=seasonal.index, y=seasonal)
fig_resid = px.line(resid, x=resid.index, y=resid)
fig_observed = px.line(observed, x=observed.index, y=observed)
# проверка на стационарность
adf_test_series = functions.adf_test(df_resample)
adf_test = adf_test_series.reset_index()
adf_test = adf_test.rename(columns={'index':'поле', 0:'значение'})
# KPSS test
kpss_test_series = functions.kpss_test(df_resample)
kpss_test = kpss_test_series.reset_index()
kpss_test = kpss_test.rename(columns={'index':'поле', 0:'значение'})
# детрендированный временной ряд
non_trend = resid + observed + seasonal
df_non_trend = non_trend.to_frame(name='num_orders')
# дифференцированный временной ряд
df_diff = df_resample.copy()
df_diff['num_orders_diff'] = df_diff['num_orders'] - df_diff['num_orders'].shift(1)
df_diff['num_orders_diff'] = df_diff['num_orders_diff'].dropna()
df_diff = df_diff.drop('num_orders', axis=1)
# конструирование признаков
df_resample['month'] = df_resample.index.month
df_resample['hour'] = df_resample.index.hour
df_resample['dayofweek'] = df_resample.index.dayofweek
df_resample['rolling_mean_D'] = df_trend['trend'].rolling(24).mean()
df_resample['rolling_mean_W'] = df_trend['trend'].rolling(168).mean()
data_months = [df_resample['2018-03-01':'2018-03-31'], df_resample['2018-04-01':'2018-04-30'], df_resample['2018-05-01':'2018-05-31'],
df_resample['2018-06-01':'2018-06-30'], df_resample['2018-07-01':'2018-07-31'], df_resample['2018-08-01':'2018-08-31']]
months = ['Март', 'Апрель', 'Май', 'Июнь', 'Июль', 'Август']
#----------------------------------------------------АНАЛИЗ------------------------------------------------------------
fig_total_trand = px.line(df_resample, x=df_resample.index, y=df_resample.rolling_mean_D, title='Общая тенденция')
fig_total_trand.add_scatter(x=df_resample.index, y=df_resample.rolling_mean_W, mode='lines', name='weekly')
# заказы по времени
hour_group = df_resample.groupby('hour').agg(func=sum).sort_values(by='num_orders', ascending=False)
fig_hour = px.bar(hour_group, x=hour_group.index, y='num_orders', hover_data=['num_orders'], color='num_orders',
labels={'num_orders': 'Заказы такси', 'hour': 'часы'}, height=400, title='Распределение заказов по времени суток')
# заказы по дням недели
weekday_group = df_resample.groupby('dayofweek').agg(func=sum)
fig_week = px.bar(weekday_group, x=weekday_group.index, y='num_orders', hover_data=['num_orders'], color='num_orders',
labels={'num_orders': 'Заказы такси', 'dayofweek': 'дни недели'},
height=400, title='Распределение заказов по дням недели')
# заказы по месяцам
fig_month = go.Figure(layout=go.Layout(template='plotly_dark'))
fig_month.add_trace(go.Box(x=data_months[0].num_orders, name='март'))
fig_month.add_trace(go.Box(x=data_months[1].num_orders, name='апрель'))
fig_month.add_trace(go.Box(x=data_months[2].num_orders, name='май'))
fig_month.add_trace(go.Box(x=data_months[3].num_orders, name='июнь'))
fig_month.add_trace(go.Box(x=data_months[4].num_orders, name='июль'))
fig_month.add_trace(go.Box(x=data_months[5].num_orders, name='август'))
fig_month.update_layout(title='Сравнение количества заказов по месяцам', xaxis_title='Количество заказов')
# вычисление автокорреляции до дифференцирования
fig_acf = functions.create_corr_plot(df_resample.num_orders, plot_pacf=False)
fig_pacf = functions.create_corr_plot(df_resample.num_orders, plot_pacf=True)
#-------------------------------------------------ОБУЧЕНИЕ и ТЕСТИРОВАНИЕ----------------------------------------------
df_resample = df_resample.dropna()
df_dammies = pd.get_dummies(df_resample, columns=['month', 'hour', 'dayofweek'], drop_first=False)
# разделение данных на обучающую и тестовую выборки
train, test = train_test_split(df_dammies, test_size=0.2, shuffle=False, random_state=12345)
#train = train.dropna()
X_train = train.drop('num_orders', axis=1)
y_train = train['num_orders']
X_test = test.drop('num_orders', axis=1)
y_test = test['num_orders']
# масштабирование данных
sc = StandardScaler()
sc.fit(X_train)
X_train_std = sc.transform(X_train)
X_test_std = sc.transform(X_test)
# моделирование и обучение
tscv = TimeSeriesSplit(n_splits=11)
models = [RidgeCV(cv=tscv),
LassoCV(cv=tscv),
ElasticNetCV(cv=tscv),
RandomForestRegressor(),
LGBMRegressor()]
pred_train = []
pred_test = []
rmse_train = []
rmse_test = []
train_graphs = []
params = []
for model in models:
# обучение
model.fit(X_train_std, y_train)
# прогноз на тренировочной выборке
prediction_train = model.predict(X_train_std)
# прогноз на тестовой выборке
prediction_test = model.predict(X_test_std)
# сохранение результатов прогнозирования
pred_train.append(prediction_train)
pred_test.append(prediction_test)
# оценка прогнозирование
rmse_train.append(round(np.sqrt(mean_squared_error(y_train, prediction_train))))
rmse_test.append(round(np.sqrt(mean_squared_error(y_test, prediction_test))))
# параметры модели
params.append(pd.DataFrame.from_dict(model.get_params(), orient='index').reset_index().rename \
(columns={'index': 'параметр', 0: 'значение'}))
# график обучения модели
train_graphs.append(functions.plot_learning_curves(model, X_train_std, y_train, cv=tscv))
# сравнительная таблица
comparison_table = pd.DataFrame({
'Model': ['RidgeCV', 'LassoCV', 'ElasticNetCV', 'RandomForestRegressor',
'LGBMRegressor'],
'RMSE train': [rmse_train[0], rmse_train[1], rmse_train[2], rmse_train[3], rmse_train[4]],
'RMSE test': [rmse_test[0], rmse_test[1], rmse_test[2], rmse_test[3], rmse_test[4]]})
# удалим параметр 'cv' из таблиц параметров
for i in range(len(params)):
params[i] = params[i][params[i]['параметр'] != 'cv']
# графики предсказаний
result_dfs = []
pred_graphs = []
for i in range(len(models)):
result_dfs.append(pd.DataFrame(pred_test[i], index=y_test.index, columns=['pred_test']).merge(y_test, on='datetime'))
fig = px.line(result_dfs[i], x=result_dfs[i].index, y=result_dfs[i].num_orders,
title='Предсказания на тестовой выборке ' + type(models[i]).__name__)
fig.add_scatter(x=result_dfs[i].index, y=result_dfs[i].pred_test, mode='lines', name='pred_test')
pred_graphs.append(fig)