class Preprocess:
def __init__(self, inputfile):
self.df = pd.read_csv(inputfile, index_col=0)
self.df['date_time'] = pd.to_datetime(self.df.date_time)
self.f = Features()
def x_features1(self):
self.df = self.f.add_bbands(self.df)
self.df = self.f.compute_side(self.df)
self.df = self.f.add_momantum(self.df)
self.df = self.f.add_volatility(self.df)
self.df = self.f.add_serial_correlation(self.df)
self.df = self.f.add_log_returns(self.df)
self.df = self.f.mov_average(self.df)
self.df = self.f.add_rsi(self.df)
self.df = self.f.add_kdj(self.df)
self.df = self.f.add_srsi(self.df)
self.df = self.f.add_cci(self.df)
self.df = self.f.add_williams(self.df)
self.df = self.f.add_trending_signal(self.df)
return self.df
"""
get side when min value touches the bbon
"""
def x_feature2(self):
self.df = self.f.add_bbands(self.df)
self.df = self.f.compute_side_min(self.df)
self.df = self.f.add_momantum(self.df)
self.df = self.f.add_volatility(self.df)
self.df = self.f.add_serial_correlation(self.df)
self.df = self.f.add_log_returns(self.df)
self.df = self.f.mov_average(self.df)
self.df = self.f.add_rsi(self.df)
self.df = self.f.add_srsi(self.df)
self.df = self.f.add_trending_signal(self.df)
return self.df
def clean_df(self):
self.df.dropna(axis=0, how='any', inplace=True)
return self.df
"""
print highly correlated pairs
"""
def correlation(self, corr_matrix=0.6):
corr_matrix = self.df.corr().abs()
high_corr_var = np.where(corr_matrix > 0.6)
high_corr_var = [(corr_matrix.columns[x], corr_matrix.columns[y])
for x, y in zip(*high_corr_var) if x != y and x < y]
print(high_corr_var)
def check_null(self):
print(self.df.isnull().sum())
sns.heatmap(self.df.isnull(), cmap="viridis")
plt.show()
def labeling(self, df=None):
fl = None
if df.shape[0] != 0:
fl = filter_label.Filter_label(df=df)
fl.cusum_filter()
else:
fl = filter_label.Filter_label(df=self.df)
fl.cusum_filter()
return fl.triple_barrier()
"""
label dataset based on vol
"""
def label_vol(self, is_infile=False, inputfile=None):
if is_infile:
df = pd.read_csv(inputfile, index_col=0)
df['date_time'] = pd.to_datetime(df.date_time)
df.index = df.date_time
df.drop(columns=['date_time'], inplace=True)
fl = filter_label.Filter_label(df=df)
fl.cusum_filter()
return fl.triple_barrier()
else:
fl = filter_label.Filter_label(df=self.df)
fl.cusum_filter()
return fl.triple_barrier()
"""
label dataset based on target
"""
def label_fix(self, target=0.01, is_infile=False, inputfile=None):
if is_infile:
df = pd.read_csv(inputfile, index_col=0)
df['date_time'] = pd.to_datetime(df.date_time)
df.index = df.date_time
df.drop(columns=['date_time'], inplace=True)
fl = filter_label.Filter_label(df=df)
fl.cusum_filter()
return fl.triple_barrier_fix(target=target)
else:
self.df['date_time'] = pd.to_datetime(self.df.date_time)
self.df.index = self.df.date_time
self.df.drop(columns=['date_time'], inplace=True)
fl = filter_label.Filter_label(df=self.df)
fl.cusum_filter()
return fl.triple_barrier_fix(target=target)
def label_fix_no_side(self, target=0.01, is_infile=False, inputfile=None):
if is_infile:
df = pd.read_csv(inputfile, index_col=0)
df['date_time'] = pd.to_datetime(df.date_time)
df.index = df.date_time
df.drop(columns=['date_time'], inplace=True)
fl = filter_label.Filter_label(df=df)
fl.cusum_filter()
return fl.triple_barrier_fix_no_side(target=target)
else:
self.df['date_time'] = pd.to_datetime(self.df.date_time)
self.df.index = self.df.date_time
self.df.drop(columns=['date_time'], inplace=True)
fl = filter_label.Filter_label(df=self.df)
fl.cusum_filter()
return fl.triple_barrier_fix_no_side(target=target)
from features import Features
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
f = Features()
df = pd.read_csv('test.csv', index_col=0)
df.date_time = pd.to_datetime(df['date_time'])
# df = f.add_bbands(df)
# df = f.compute_side(df)
k = f.add_kdj(df)
# print(df.columns)
# print(type(k.indx))
# df = f.add_momantum(df)
# df = f.add_volatility(df)
# df = f.add_serial_correlation(df)
# df = f.add_log_returns(df)
# df = f.mov_average(df)
# df = f.add_trending_signal(df)
df.to_csv('demo.csv')