| from pathlib import Path
|
| from typing import Union
|
| import json
|
|
|
| import pandas as pd
|
|
|
| PACKAGE_ROOT = Path(__file__).parent.parent
|
| #PACKAGE_PARENT = '..'
|
| #SCRIPT_DIR = os.path.dirname(os.path.realpath(os.path.join(os.getcwd(), os.path.expanduser(__file__))))
|
| #sys.path.append(os.path.normpath(os.path.join(SCRIPT_DIR, PACKAGE_PARENT)))
|
| #PACKAGE_ROOT = os.path.dirname(os.path.abspath(__file__))
|
|
|
|
|
| class App:
|
| """Globally visible variables."""
|
|
|
| # System
|
| loop = None # asyncio main loop
|
| sched = None # Scheduler
|
|
|
| analyzer = None # Store and analyze data
|
|
|
| # Connector client
|
| client = None
|
|
|
| # WebSocket for push notifications
|
| bm = None
|
| conn_key = None # Socket
|
|
|
| signal = None, # Latest signal "BUY", "SELL"
|
|
|
| #
|
| # State of the server (updated after each interval)
|
| #
|
| # State 0 or None or empty means ok. String and other non empty objects mean error
|
| error_status = 0 # Networks, connections, exceptions etc. what does not allow us to work at all
|
| server_status = 0 # If server allow us to trade (maintenance, down etc.)
|
| account_status = 0 # If account allows us to trade (funds, suspended etc.)
|
| trade_state_status = 0 # Something wrong with our trading logic (wrong use, inconsistent state etc. what we cannot recover)
|
|
|
| # Trade status
|
| transaction = None
|
| status = None # BOUGHT, SOLD, BUYING, SELLING
|
| order = None # Latest or current order
|
| order_time = None # Order submission time
|
|
|
| # Available assets for trade
|
| # Can be set by the sync/recover function or updated by the trading algorithm
|
| base_quantity = "0.04108219" # BTC owned (on account, already bought, available for trade)
|
| quote_quantity = "1000.0" # USDT owned (on account, available for trade)
|
|
|
| #
|
| # Trader. Status data retrieved from the server. Below are examples only.
|
| #
|
| system_status = {"status": 0, "msg": "normal"} # 0: normal,1:system maintenance
|
| symbol_info = {}
|
| account_info = {}
|
|
|
| #
|
| # Constant configuration parameters
|
| #
|
| config = {
|
| "actions": ["notify"], # Values: notify, trade
|
|
|
| # Binance
|
| "api_key": "",
|
| "api_secret": "",
|
|
|
| # Telegram
|
| "telegram_bot_token": "", # Source address of messages
|
| "telegram_chat_id": "", # Destination address of messages
|
|
|
| # File locations
|
| "data_folder": "", # It is needed for model training
|
| # Symbol determines sub-folder and used in other identifiers
|
| "symbol": "", # BTCUSDT ETHUSDT
|
| # It will be appended to generated file names and denotes config (like config name).
|
| # TODO: In future, we could/should change it semantics, and store the generated files in the folder with this name, rather than append a modifier to file names
|
| "config_file_modifier": "",
|
|
|
| # File name conventions
|
| "merge_file_modifier": "data",
|
| "feature_file_modifier": "features",
|
| "matrix_file_modifier": "matrix",
|
| "predict_file_modifier": "predictions", # predict, predict-rolling
|
| "signal_file_modifier": "performance",
|
|
|
| # =====================================
|
| # === MERGER and (online) COLLECTOR ===
|
|
|
| # Specify which data sources to merge into one file with all source columns
|
| # Format: (folder, file, prefix) - (symbol, data source, column prefix)
|
| # These descriptors are also used to retrieve data by collector in online mode - folder name is symbol name
|
| #"data_sources": [
|
| # {"folder": "BTCUSDT", "file": "klines", "column_prefix": ""},
|
| #],
|
| #"data_sources": [
|
| # {"folder": "BTCUSDT", "file": "klines", "column_prefix": "btc"},
|
| # {"folder": "ETHUSDT", "file": "klines", "column_prefix": "eth"},
|
| #],
|
|
|
| # ==========================
|
| # === FEATURE GENERATION ===
|
|
|
| # What columns to pass to which feature generator and how to prefix its derived features
|
| "feature_sets": [
|
| {"column_prefix": "btc", "generator": "klines", "feature_prefix": "btc"},
|
| {"column_prefix": "eth", "generator": "klines", "feature_prefix": "eth"},
|
| ],
|
| # Parameters of klines feature generator
|
| # If these are changed, then feature names (below) will also have to be changed
|
|
|
| #"base_window_kline": 40320, # 4 weeks
|
| #"windows_kline": [1, 20, 60, 360, 1440, 10080],
|
| #"area_windows_kline": [20, 60, 360, 1440, 10080],
|
|
|
| #"base_window_kline": 20160, # 2 weeks
|
| #"windows_kline": [1, 20, 60, 180, 720, 2880],
|
| #"area_windows_kline": [60, 120, 180, 720, 2880],
|
| #"windows_kline": [1, 30, 120, 360, 1440, 4320],
|
| #"area_windows_kline": [30, 120, 360, 1440, 4320],
|
| #"windows_kline": [1, 60, 180, 720, 2880, 5760],
|
| #"area_windows_kline": [60, 180, 720, 2880, 5760],
|
|
|
| "base_window_kline": 10080, # 1 week
|
| "windows_kline": [1, 10, 30, 180, 720, 1440],
|
| "area_windows_kline": [10, 30, 180, 720, 1440],
|
|
|
| # ========================
|
| # === LABEL GENERATION ===
|
|
|
| "high_low_horizon": 1440, # Parameter of labels: computing max and min for this horizon ahead
|
| "top_bot_column_name": "close",
|
|
|
| # ===========================
|
| # === MODEL TRAIN/PREDICT ===
|
| # off-line and on-line
|
|
|
| # Minimum history required to compute derived features
|
| # It is used in online mode where we need to maintain data window of this size or larger
|
| # Take maximum aggregation windows from feature generation code (and add something to be sure that we have all what is needed)
|
| # Basically, should be equal to base_window_kline
|
| "features_horizon": 10180,
|
|
|
| # One model is for each (label, train_features, algorithm)
|
|
|
| # Feature column names returned by the klines feature generator
|
| # They are used by train/predict
|
| "features_kline": [
|
| 'btc_close_1', 'btc_close_10', 'btc_close_30', 'btc_close_180', 'btc_close_720', 'btc_close_1440',
|
| 'btc_close_std_10', 'btc_close_std_30', 'btc_close_std_180', 'btc_close_std_720', 'btc_close_std_1440',
|
| 'btc_volume_1', 'btc_volume_10', 'btc_volume_30', 'btc_volume_180', 'btc_volume_720', 'btc_volume_1440',
|
| 'btc_span_1', 'btc_span_10', 'btc_span_30', 'btc_span_180', 'btc_span_720', 'btc_span_1440',
|
| 'btc_trades_1', 'btc_trades_10', 'btc_trades_30', 'btc_trades_180', 'btc_trades_720', 'btc_trades_1440',
|
| 'btc_tb_base_1', 'btc_tb_base_10', 'btc_tb_base_30', 'btc_tb_base_180', 'btc_tb_base_720', 'btc_tb_base_1440',
|
| #'btc_close_area_10', 'btc_close_area_30', 'btc_close_area_180', 'btc_close_area_720', 'btc_close_area_1440',
|
| 'btc_close_trend_10', 'btc_close_trend_30', 'btc_close_trend_180', 'btc_close_trend_720', 'btc_close_trend_1440',
|
| 'btc_volume_trend_10', 'btc_volume_trend_30', 'btc_volume_trend_180', 'btc_volume_trend_720', 'btc_volume_trend_1440',
|
|
|
| 'eth_close_1', 'eth_close_10', 'eth_close_30', 'eth_close_180', 'eth_close_720', 'eth_close_1440',
|
| 'eth_close_std_10', 'eth_close_std_30', 'eth_close_std_180', 'eth_close_std_720', 'eth_close_std_1440',
|
| 'eth_volume_1', 'eth_volume_10', 'eth_volume_30', 'eth_volume_180', 'eth_volume_720', 'eth_volume_1440',
|
| 'eth_span_1', 'eth_span_10', 'eth_span_30', 'eth_span_180', 'eth_span_720', 'eth_span_1440',
|
| 'eth_trades_1', 'eth_trades_10', 'eth_trades_30', 'eth_trades_180', 'eth_trades_720', 'eth_trades_1440',
|
| 'eth_tb_base_1', 'eth_tb_base_10', 'eth_tb_base_30', 'eth_tb_base_180', 'eth_tb_base_720', 'eth_tb_base_1440',
|
| #'eth_close_area_10', 'eth_close_area_30', 'eth_close_area_180', 'eth_close_area_720', 'eth_close_area_1440',
|
| 'eth_close_trend_10', 'eth_close_trend_30', 'eth_close_trend_180', 'eth_close_trend_720', 'eth_close_trend_1440',
|
| 'eth_volume_trend_10', 'eth_volume_trend_30', 'eth_volume_trend_180', 'eth_volume_trend_720', 'eth_volume_trend_1440'
|
| ],
|
|
|
| # Used to select columns for training by adding together different feature sets (the lists are defined elswhere in variables like "features_kline")
|
| "train_features": ["kline"],
|
| # algorithm descriptors from model store
|
| "algorithms": ["nn"], # gb, nn, lc - definitions of their parameters are in model store
|
|
|
| # This will be excluded from model training
|
| "label_horizon": 0,
|
|
|
| # These labels are specified manually and are produced by label generator
|
| # Models will be trained for these models
|
| "labels": [
|
| "bot2_025", "bot2_05", "bot2_075", "bot2_1", "bot2_125", "bot2_15",
|
| "top2_025", "top2_05", "top2_075", "top2_1", "top2_125", "top2_15",
|
| ],
|
| "_labels": [
|
| "bot3_025", "bot3_05", "bot3_075", "bot3_1", "bot3_125", "bot3_15", "bot3_175",
|
| "top3_025", "top3_05", "top3_075", "top3_1", "top3_125", "top3_15", "top3_175",
|
|
|
| "bot4_025", "bot4_05", "bot4_075", "bot4_1", "bot4_125", "bot4_15", "bot4_175", "bot4_2",
|
| "top4_025", "top4_05", "top4_075", "top4_1", "top4_125", "top4_15", "top4_175", "top4_2",
|
|
|
| "bot5_025", "bot5_05", "bot5_075", "bot5_1", "bot5_125", "bot5_15", "bot5_175", "bot5_2", "bot5_25",
|
| "top5_025", "top5_05", "top5_075", "top5_1", "top5_125", "top5_15", "top5_175", "top5_2", "top5_25",
|
|
|
| 'high_max_180', # Maximum high (relative)
|
|
|
| "high_10", "high_15", "high_20", "high_25", "high_30",
|
| "low_10", "low_15", "low_20", "low_25", "low_30",
|
|
|
| 'low_min_180', # Minimum low (relative)
|
| 'low_01', 'low_02', 'low_03', 'low_04', # Always above
|
| 'low_10', 'low_15', 'low_20', 'low_25', # At least one time below
|
|
|
| 'high_to_low_180',
|
|
|
| 'close_area_future_60', 'close_area_future_120', 'close_area_future_180', 'close_area_future_300',
|
| ],
|
|
|
| # =========================
|
| # === SIGNAL GENERATION ===
|
|
|
| # These are predicted columns <label, train_features, algorithm> as well as model (pair) names
|
| "buy_labels": [],
|
| "sell_labels": [],
|
| "_buy_labels": ["bot4_1_k_lc", "bot4_15_k_lc", "bot4_2_k_lc", "bot4_25_k_lc", "bot4_3_k_lc"],
|
| "_sell_labels": ["top4_1_k_lc", "top4_15_k_lc", "top4_2_k_lc", "top4_25_k_lc", "top4_3_k_lc"],
|
|
|
| # It defines how signal scores, trade signals, and notification signals will be generated
|
| # from point-wise prediction scores for two groups of labels
|
| "signal_model": {
|
| # First, aggregation in group over various algorithms and label parameters
|
| "buy_point_threshold": None, # Second, produce boolean column (optional)
|
| "buy_window": 3, # Third, aggregate in time
|
| # Now we have the final score
|
| "buy_signal_threshold": 0.65, # To decide whether to buy/sell after all aggregations/combinations
|
| "buy_notify_threshold": 0.05, # To decide whether to notify (can be an option of individual users/consumers)
|
|
|
| "combine": "", # "no_combine", "relative", "difference" Find relative/difference
|
|
|
| "sell_point_threshold": None,
|
| "sell_window": 3,
|
| "sell_signal_threshold": 0.65,
|
| "sell_notify_threshold": 0.05,
|
|
|
| "trade_icon_step": 0.1, # For each step, one icon added
|
| "notify_frequency_minutes": 10, # 1m, 5m, 10m, 15m etc. Minutes will be divided by this number
|
| },
|
|
|
| # =====================
|
| # === TRADER SERVER ===
|
| "base_asset": "", # BTC ETH
|
| "quote_asset": "",
|
|
|
| "trader": {
|
| # For debugging: determine what parts of code will be executed
|
| "no_trades_only_data_processing": False, # in market or out of market processing is excluded (all below parameters ignored)
|
| "test_order_before_submit": False, # Send test submit to the server as part of validation
|
| "simulate_order_execution": False, # Instead of real orders, simulate their execution (immediate buy/sell market orders and use high price of klines for limit orders)
|
|
|
| "percentage_used_for_trade": 99, # in % to the available USDT quantity, that is, we will derive how much BTC to buy using this percentage
|
| "limit_price_adjustment": -0.0001, # Limit price of orders will be better than the latest close price (0 means no change, positive - better for us, negative - worse for us)
|
|
|
| # Signal model (trade strategy) - currently NOT USED
|
| "sell_timeout": 70, # Seconds
|
| "percentage_sell_price": 1.018, # our planned profit per trade via limit sell order (part of the model)
|
| },
|
|
|
| # ==================
|
| # === COLLECTORS ===
|
| "collector": {
|
| "folder": "DATA",
|
| "flush_period": 300, # seconds
|
| "depth": {
|
| "folder": "DEPTH",
|
| "symbols": ["BTCUSDT", "ETHBTC", "ETHUSDT", "IOTAUSDT", "IOTABTC", "IOTAETH"],
|
| "limit": 100, # Legal values (depth): '5, 10, 20, 50, 100, 500, 1000, 5000' <100 weight=1
|
| "freq": "1m", # Binance standard frequency: 5s, 1m etc.
|
| },
|
| "stream": {
|
| "folder": "STREAM",
|
| # Stream formats:
|
| # For kline channel: <symbol>@kline_<interval>, Event type: "e": "kline", Symbol: "s": "BNBBTC"
|
| # For depth channel: <symbol>@depth<levels>[@100ms], Event type: NO, Symbol: NO
|
| # btcusdt@ticker
|
| "channels": ["kline_1m", "depth20"], # kline_1m, depth20, depth5
|
| "symbols": ["BTCUSDT", "ETHBTC", "ETHUSDT", "IOTAUSDT", "IOTABTC", "IOTAETH"],
|
| # "BTCUSDT", "ETHBTC", "ETHUSDT", "IOTAUSDT", "IOTABTC", "IOTAETH"
|
| }
|
| },
|
| }
|
|
|
|
|
| def data_provider_problems_exist():
|
| if App.error_status != 0:
|
| return True
|
| if App.server_status != 0:
|
| return True
|
| return False
|
|
|
|
|
| def problems_exist():
|
| if App.error_status != 0:
|
| return True
|
| if App.server_status != 0:
|
| return True
|
| if App.account_status != 0:
|
| return True
|
| if App.trade_state_status != 0:
|
| return True
|
| return False
|
|
|
|
|
| def load_config(config_file):
|
| if config_file:
|
| config_file_path = PACKAGE_ROOT / config_file
|
| with open(config_file_path) as json_file:
|
| config_json = json.load(json_file)
|
| App.config.update(config_json)
|
|
|
|
|
| if __name__ == "__main__":
|
| pass
|