Update MLCryptoForecasterAllAssetsTPSL_ParisTime.py

MLCryptoForecasterAllAssetsTPSL_ParisTime.py

@@ -1,6 +1,7 @@
 import os
 import pandas as pd
 import numpy as np
+import argparse
 from datetime import timedelta
 from binance.client import Client
 from sklearn.model_selection import train_test_split
@@ -9,16 +10,34 @@ from sklearn.metrics import classification_report
 import ta
 import pytz
 
+# Parse command-line arguments for timeframe
+parser = argparse.ArgumentParser(description="Binance Trend Forecaster with adjustable timeframe")
+parser.add_argument("--interval", type=str, default="4h",
+                    choices=["1m","3m","5m","15m","30m","1h","4h","1d"],
+                    help="Time interval for klines (e.g. '1h', '4h', '1d')")
+args = parser.parse_args()
+
+# Map user-friendly intervals to Binance API constants
+interval_map = {
+    "1m": Client.KLINE_INTERVAL_1MINUTE,
+    "3m": Client.KLINE_INTERVAL_3MINUTE,
+    "5m": Client.KLINE_INTERVAL_5MINUTE,
+    "15m": Client.KLINE_INTERVAL_15MINUTE,
+    "30m": Client.KLINE_INTERVAL_30MINUTE,
+    "1h": Client.KLINE_INTERVAL_1HOUR,
+    "4h": Client.KLINE_INTERVAL_4HOUR,
+    "1d": Client.KLINE_INTERVAL_1DAY
+}
+interval = interval_map[args.interval]
+
 # Function to log results to both console and file
 # Blank lines are added after each asset block explicitly
-
 def log_results(message, filename="predictions_results.txt"):
     print(message)
     with open(filename, "a") as f:
         f.write(message + "\n")
 
 # Convert UTC timestamp to Europe/Paris timezone
-
 def convert_to_paris_time(utc_time):
     paris_tz = pytz.timezone('Europe/Paris')
     utc_time = utc_time.replace(tzinfo=pytz.utc)
@@ -29,8 +48,7 @@ def convert_to_paris_time(utc_time):
 client = Client()
 
 # Settings
-interval = Client.KLINE_INTERVAL_4HOUR
-result_file = "predictions_results.txt"
+result_file = f"predictions_results_{args.interval}.txt"
 
 # Delete the results file if it exists for a fresh start
 if os.path.exists(result_file):
@@ -67,14 +85,16 @@ def optimize_tp_sl(df, signals, side, pgrid, lgrid):
 
 # Main loop: process each symbol
 for symbol in symbols:
-    log_results(f"=== {symbol} ===", result_file)
+    log_results(f"=== {symbol} ({args.interval}) ===", result_file)
 
     # Load or download historical data
-    data_file = f"{symbol}_data_4h_full.csv"
+    data_file = f"{symbol}_data_{args.interval}_full.csv"
     if os.path.exists(data_file):
         df = pd.read_csv(data_file, index_col=0, parse_dates=True)
         last_ts = df.index[-1]
-        start = (last_ts + timedelta(hours=4)).strftime("%d %B %Y %H:%M:%S")
+        start = (last_ts + timedelta(**{
+            'minutes':1 if args.interval=='1m' else 3 if args.interval=='3m' else 5 if args.interval=='5m' else 15 if args.interval=='15m' else 30 if args.interval=='30m' else 60 if args.interval=='1h' else 240 if args.interval=='4h' else 1440
+        })).strftime("%d %B %Y %H:%M:%S")
         new = client.get_historical_klines(symbol, interval, start)
         if new:
             new_df = pd.DataFrame(new, columns=[
@@ -120,10 +140,10 @@ for symbol in symbols:
     df.dropna(inplace=True)
 
     # Label signals based on Ichimoku cloud
-    df['signal'] = np.select(
-        [(df['close'] > df['span_a']) & (df['close'] > df['span_b']),
-         (df['close'] < df['span_a']) & (df['close'] < df['span_b'])],
-        [1, 0], default=-1)
+    df['signal'] = np.select([
+        (df['close'] > df['span_a']) & (df['close'] > df['span_b']),
+        (df['close'] < df['span_a']) & (df['close'] < df['span_b'])
+    ], [1, 0], default=-1)
 
     # Train/test split
     features = [c for c in df.columns if c not in ['open','high','low','close','volume','signal']]
@@ -162,4 +182,4 @@ for symbol in symbols:
         f.write("\n")
 
 # End of processing
-log_results("All assets processed.", result_file)
+log_results("All assets processed.", result_file)