352 lines
15 KiB
Python
352 lines
15 KiB
Python
import os
| |
from datetime import datetime, timezone, timedelta
| |
from typing import Union
| |
import json
| |
| |
import numpy as np
| |
import pandas as pd
| |
| |
from common.utils import *
| |
| |
"""
| |
Feature/label generation.
| |
These features are computed using explict transformations.
| |
(True) labels are features computed from future data but stored as properties of the current row (in contrast to normal features which are computed from past data).
| |
(Currently) feature/label generation is not based on (explicit) models - all parameters are hard-coded.
| |
Also, no parameter training is performed.
| |
"""
| |
| |
| |
def generate_features(df,use_differences, base_window, windows, area_windows):
| |
"""
| |
Generate derived features by adding them as new columns to the data frame.
| |
It is important that the same parameters are used for both training and prediction.
| |
| |
Most features compute rolling aggregation. However, instead of absolute values, the difference
| |
of this rolling aggregation to the (longer) base rolling aggregation is computed.
| |
| |
The window sizes are used for encoding feature/column names and might look like 'close_120'
| |
for average close price for the last 120 minutes (relative to the average base price).
| |
The column names are needed when preparing data for training or prediction.
| |
The easiest way to get them is to return from this function and copy and the
| |
corresponding config attribute.
| |
"""
| |
features = []
| |
to_drop = []
| |
| |
if use_differences:
| |
df['close'] = to_diff(df['close'])
| |
df['volume'] = to_diff(df['volume'])
| |
df['trades'] = to_diff(df['trades'])
| |
| |
# close mean
| |
weight_column_name = 'volume' # None: no weighting; 'volume': volume average
| |
to_drop += add_past_weighted_aggregations(df, 'close', weight_column_name, np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_weighted_aggregations(df, 'close', weight_column_name, np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['close_1', 'close_2', 'close_5', 'close_20', 'close_60', 'close_180']
| |
| |
# close std
| |
to_drop += add_past_aggregations(df, 'close', np.nanstd, base_window) # Base column
| |
features += add_past_aggregations(df, 'close', np.nanstd, windows[1:], '_std', to_drop[-1], 100.0) # window 1 excluded
| |
# ['close_std_1', 'close_std_2', 'close_std_5', 'close_std_20', 'close_std_60', 'close_std_180']
| |
| |
# volume mean
| |
to_drop += add_past_aggregations(df, 'volume', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'volume', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['volume_1', 'volume_2', 'volume_5', 'volume_20', 'volume_60', 'volume_180']
| |
| |
# Span: high-low difference
| |
df['span'] = df['high'] - df['low']
| |
to_drop.append('span')
| |
to_drop += add_past_aggregations(df, 'span', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'span', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['span_1', 'span_2', 'span_5', 'span_20', 'span_60', 'span_180']
| |
| |
# Number of trades
| |
to_drop += add_past_aggregations(df, 'trades', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'trades', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['trades_1', 'trades_2', 'trades_5', 'trades_20', 'trades_60', 'trades_180']
| |
| |
# tb_base_av / volume varies around 0.5 in base currency
| |
df['tb_base'] = df['tb_base_av'] / df['volume']
| |
to_drop.append('tb_base')
| |
to_drop += add_past_aggregations(df, 'tb_base', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'tb_base', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['tb_base_1', 'tb_base_2', 'tb_base_5', 'tb_base_20', 'tb_base_60', 'tb_base_180']
| |
| |
# tb_quote_av / quote_av varies around 0.5 in quote currency
| |
df['tb_quote'] = df['tb_quote_av'] / df['quote_av']
| |
to_drop.append('tb_quote')
| |
to_drop += add_past_aggregations(df, 'tb_quote', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'tb_quote', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['tb_quote_1', 'tb_quote_2', 'tb_quote_5', 'tb_quote_20', 'tb_quote_60', 'tb_quote_180']
| |
| |
# Area over and under latest close price
| |
features += add_area_ratio(df, is_future=False, column_name="close", windows=area_windows, suffix = "_area")
| |
| |
# Linear trend
| |
features += add_linear_trends(df, is_future=False, column_name="close", windows=windows[1:], suffix="_trend") # window 1 excluded
| |
| |
df.drop(columns=to_drop, inplace=True)
| |
| |
return features
| |
| |
| |
def generate_features_futur(df, use_differences=False):
| |
"""
| |
Generate derived features for futures.
| |
"""
| |
# Parameters of moving averages
| |
windows = [1, 2, 5, 20, 60, 180]
| |
base_window = 360
| |
| |
features = []
| |
to_drop = []
| |
| |
if use_differences:
| |
df['f_close'] = to_diff(df['f_close'])
| |
df['f_volume'] = to_diff(df['f_volume'])
| |
df['f_trades'] = to_diff(df['f_trades'])
| |
| |
# close mean
| |
weight_column_name = 'f_volume' # None: no weighting; 'volume': volume average
| |
to_drop += add_past_weighted_aggregations(df, 'f_close', weight_column_name, np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_weighted_aggregations(df, 'f_close', weight_column_name, np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['f_close_1', f_close_2', 'f_close_5', 'f_close_10', 'f_close_20']
| |
| |
# close std
| |
to_drop += add_past_aggregations(df, 'f_close', np.nanstd, base_window) # Base column
| |
features += add_past_aggregations(df, 'f_close', np.nanstd, windows[1:], '_std', to_drop[-1], 100.0) # window 1 excluded
| |
# ['f_close_std_1', f_close_std_2', 'f_close_std_5', 'f_close_std_10', 'f_close_std_20']
| |
| |
# volume mean
| |
to_drop += add_past_aggregations(df, 'f_volume', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'f_volume', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['f_volume_1', 'f_volume_2', 'f_volume_5', 'f_volume_10', 'f_volume_20']
| |
| |
# Span: high-low difference
| |
df['f_span'] = df['f_high'] - df['f_low']
| |
to_drop.append('f_span')
| |
to_drop += add_past_aggregations(df, 'f_span', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'f_span', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['f_span_1', 'f_span_2', 'f_span_5', 'f_span_10', 'f_span_20']
| |
| |
# Number of trades
| |
to_drop += add_past_aggregations(df, 'f_trades', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'f_trades', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['f_trades_1', 'f_trades_2', 'f_trades_5', 'f_trades_10', 'f_trades_20']
| |
| |
# tb_base_av / volume varies around 0.5 in base currency
| |
#df['f_tb_base'] = df['f_tb_base_av'] / df['f_volume']
| |
#to_drop.append('f_tb_base')
| |
#to_drop += add_past_aggregations(df, 'f_tb_base', np.nanmean, base_window, suffix='') # Base column
| |
#features += add_past_aggregations(df, 'f_tb_base', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['f_tb_base_1', 'f_tb_base_2', 'f_tb_base_5', 'f_tb_base_10', 'f_tb_base_20']
| |
| |
# tb_quote_av / quote_av varies around 0.5 in quote currency
| |
#df['f_tb_quote'] = df['f_tb_quote_av'] / df['f_quote_av']
| |
#to_drop.append('f_tb_quote')
| |
#to_drop += add_past_aggregations(df, 'f_tb_quote', np.nanmean, base_window, suffix='') # Base column
| |
#features += add_past_aggregations(df, 'f_tb_quote', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['f_tb_quote_1', 'f_tb_quote_2', 'f_tb_quote_5', 'f_tb_quote_10', 'f_tb_quote_20']
| |
| |
# Area over and under latest close price
| |
features += add_area_ratio(df, is_future=False, column_name="f_close", windows=[20, 60, 120, 180], suffix = "_area")
| |
| |
# Linear trend
| |
features += add_linear_trends(df, is_future=False, column_name="f_close", windows=windows[1:], suffix="_trend") # window 1 excluded
| |
| |
df.drop(columns=to_drop, inplace=True)
| |
| |
return features
| |
| |
| |
def generate_features_depth(df, use_differences=False):
| |
"""
| |
Generate derived features from depth data.
| |
Original features:
| |
- gap, price,
| |
- bids_1,asks_1,
| |
- bids_2,asks_2,
| |
- bids_5,asks_5,
| |
- bids_10,asks_10,
| |
- bids_20,asks_20
| |
| |
Features (33):
| |
gap_2,gap_5,gap_10,
| |
bids_1_2,bids_1_5,bids_1_10, asks_1_2,asks_1_5,asks_1_10,
| |
bids_2_2,bids_2_5,bids_2_10, asks_2_2,asks_2_5,asks_2_10,
| |
bids_5_2,bids_5_5,bids_5_10, asks_5_2,asks_5_5,asks_5_10,
| |
bids_10_2,bids_10_5,bids_10_10, asks_10_2,asks_10_5,asks_10_10,
| |
bids_20_2,bids_20_5,bids_20_10, asks_20_2,asks_20_5,asks_20_10,
| |
"""
| |
# Parameters of moving averages
| |
windows = [2, 5, 10]
| |
base_window = 30
| |
| |
features = []
| |
to_drop = []
| |
| |
# gap mean
| |
to_drop += add_past_aggregations(df, 'gap', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'gap', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['gap_2', 'gap_5', 'gap_10']
| |
| |
| |
# bids_1 mean
| |
to_drop += add_past_aggregations(df, 'bids_1', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'bids_1', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['bids_1_2', 'bids_1_5', 'bids_1_10']
| |
# asks_1 mean
| |
to_drop += add_past_aggregations(df, 'asks_1', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'asks_1', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['asks_1_2', 'asks_1_5', 'asks_1_10']
| |
| |
| |
# bids_2 mean
| |
to_drop += add_past_aggregations(df, 'bids_2', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'bids_2', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['bids_2_2', 'bids_2_5', 'bids_2_10']
| |
# asks_2 mean
| |
to_drop += add_past_aggregations(df, 'asks_2', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'asks_2', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['asks_2_2', 'asks_2_5', 'asks_2_10']
| |
| |
| |
# bids_5 mean
| |
to_drop += add_past_aggregations(df, 'bids_5', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'bids_5', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['bids_5_2', 'bids_5_5', 'bids_5_10']
| |
# asks_5 mean
| |
to_drop += add_past_aggregations(df, 'asks_5', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'asks_5', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['asks_5_2', 'asks_5_5', 'asks_5_10']
| |
| |
| |
# bids_10 mean
| |
to_drop += add_past_aggregations(df, 'bids_10', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'bids_10', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['bids_10_2', 'bids_10_5', 'bids_10_10']
| |
# asks_10 mean
| |
to_drop += add_past_aggregations(df, 'asks_10', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'asks_10', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['asks_10_2', 'asks_10_5', 'asks_10_10']
| |
| |
| |
# bids_20 mean
| |
to_drop += add_past_aggregations(df, 'bids_20', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'bids_20', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['bids_20_2', 'bids_20_5', 'bids_20_10']
| |
# asks_20 mean
| |
to_drop += add_past_aggregations(df, 'asks_20', np.nanmean, base_window, suffix='') # Base column
| |
features += add_past_aggregations(df, 'asks_20', np.nanmean, windows, '', to_drop[-1], 100.0)
| |
# ['asks_20_2', 'asks_20_5', 'asks_20_10']
| |
| |
| |
df.drop(columns=to_drop, inplace=True)
| |
| |
return features
| |
| |
| |
def depth_to_df(depth: list):
| |
"""
| |
Input is a list of json objects each representing current market depth with a list bids and asks
| |
The method computes features from the market depth and returns a data frame with the corresponding columns.
| |
| |
NOTE:
| |
- Important: "timestamp" is real time of the depth data which corresponds to "close_time" in klines
| |
"timestamp" in klines is 1m before current time
| |
It has to be taken into account when matching/joining records, e.g., by shifting columns (if we match "timestamp" then the reslt will be wrong)
| |
- data frame index is continuous and may contain gaps. its start is first line and end is last line
| |
| |
# TODO Questions:
| |
# !!! - what is zone for our timestamps - ensure that it is the same as Binance server
| |
# - is it possible to create a data frame with a column containing json object or string?
| |
# - how to match json/string values with data frame index?
| |
"""
| |
bin_size = 1.0 # In USDT
| |
windows = [1, 2, 5, 10, 20] # No of price bins for aggregate/smoothing
| |
| |
#
| |
# Generate a table with feature records
| |
#
| |
table = []
| |
for entry in depth:
| |
record = depth_to_features(entry, windows, bin_size)
| |
table.append(record)
| |
| |
#
| |
# Convert json table to data frame
| |
#
| |
df = pd.DataFrame.from_dict(table)
| |
# Alternatively, from_records() or json_normalize()
| |
| |
# Timestamp is an index
| |
df["timestamp"] = pd.to_datetime(df["timestamp"], unit='ms')
| |
df = df.set_index("timestamp")
| |
df = df.sort_index()
| |
| |
#
| |
# Find start and end dates
| |
#
| |
# NOTE: timestamp is request time (in our implementation) and hence it is end of 1m interval while kline id is start of 1m inteval
| |
# It is important for matching, so maybe align this difference here by shifting data
| |
start_line = depth[0]
| |
end_line = depth[-1]
| |
start_ts = start_line.get("timestamp")
| |
#start_ts -= 60_000 # To ensure that we do not lose any data
| |
end_ts = end_line.get("timestamp")
| |
#end_ts += 60_000 # To ensure that we do not lose any data
| |
| |
#
| |
# Create index for this interval of timestamps
| |
#
| |
# NOTE: Add utc=True to get tz-aware object (with tz="UTC" instead of tz-unaware object with tz=None), so it seems that no tz means UTC
| |
start = pd.to_datetime(start_ts, unit='ms')
| |
end = pd.to_datetime(end_ts, unit='ms')
| |
| |
# Alternatively:
| |
# If tz is not specified then 1 hour difference will be added so it seems that no tz means locale tz
| |
#datetime.fromtimestamp(float(start_ts) / 1e3, tz=pytz.UTC)
| |
| |
# Create DatetimeIndex
| |
# NOTE: if tz is not specified then the index is tz-naive
| |
# closed can be specified (which side to include/exclude: left, right or both). it influences if we want ot include/exclude start or end of the interval
| |
index = pd.date_range(start, end, freq="T")
| |
df_out = pd.DataFrame(index=index)
| |
| |
#
| |
# Join data with this empty index (to ensure continuous range of timestamps)
| |
#
| |
df_out = df_out.join(df)
| |
| |
return df_out
| |
| |
| |
def depth_to_features(entry: list, windows: list, bin_size: float):
| |
"""Convert one record of market depth to a dict of features"""
| |
| |
bids = entry.get("bids")
| |
asks = entry.get("asks")
| |
| |
timestamp = entry.get("timestamp")
| |
| |
# Gap feature
| |
gap = asks[0][0] - bids[0][0]
| |
| |
if gap < 0: gap = 0
| |
| |
# Price feature
| |
price = bids[0][0] + (gap / 2)
| |
| |
# Densities for bids and asks (volume per price unit)
| |
densities = mean_volumes(depth=entry, windows=windows, bin_size=bin_size)
| |
| |
record = {"timestamp": timestamp, "gap": gap, "price": price}
| |
record.update(densities)
| |
| |
return record
| |
| |
| |
if __name__ == "__main__":
| |
pass
|