cryptoPredict

A Python-based Bitcoin trading research project built around Binance 15-minute market data, technical indicators, model training experiments, signal generation, Telegram alerts, and Binance Futures testnet execution.

This public candidate is prepared as a safe showcase version of the original private working project. The repo intentionally keeps the code and workflow visible while omitting local credentials, virtual environments, raw datasets, trained model files, and runtime state.

Highlights

• historical spot and futures data collection
• multiple training paths including XGBoost and LSTM
• threshold tuning and periodic simulation scripts
• live signal checking from recent candles
• Telegram notification helper
• optional Binance Futures testnet execution helper

Project layout

Data collection

• data.py - download spot BTCUSDT candles and build feature-ready historical data
• mining_premium.py - collect futures candle data and funding-rate history for the premium strategy path

Training and experimentation

• latih_model.py - LSTM training pipeline
• latih_xgboost.py - XGBoost classification experiment
• latih_swing_profit.py - swing-strategy training and profit simulation
• latih_sniper.py - short-window threshold tuning for sniper-style entries
• latih_premium.py - funding-rate-aware premium model training
• latih_final.py - market behavior scan for target / holding exploration
• cari_setingan_terbaik.py - parameter search helper
• simulasi_berkala.py - grouped simulation report generator

Live helpers

• cek_sinyal.py - evaluate recent candles and produce a live decision signal
• bot_otomatis.py - 15-minute scheduler wrapper
• telegram_utils.py - Telegram notification helper using env vars
• eksekusi_demo.py - Binance Futures testnet execution helper using env vars
• cek_gpu.py - environment / accelerator check

Suggested workflow

1. collect local data with data.py or mining_premium.py
2. train one or more candidate models with the relevant latih_*.py scripts
3. keep generated artifacts locally
4. run cek_sinyal.py against fresh market data
5. optionally schedule checks with bot_otomatis.py
6. optionally forward alerts to Telegram or route to Binance Futures testnet helpers

Quick start

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env

Required local variables:

• TELEGRAM_BOT_TOKEN
• TELEGRAM_CHAT_ID
• BINANCE_TESTNET_API_KEY
• BINANCE_TESTNET_SECRET_KEY

Example commands:

python data.py
python latih_swing_profit.py
python cek_sinyal.py
python bot_otomatis.py

Documentation

• docs/QUICKSTART.md
• docs/SCRIPT_MAP.md
• docs/WORKFLOW.md

Public packaging notes

This showcase copy intentionally excludes:

• raw datasets
• trained model artifacts and scalers
• local virtualenv folders
• active runtime state
• private Telegram / Binance credentials

Notes

This repo is presented as a practical research and automation project, not financial advice. It is best understood as an evolving experiment in feature engineering, filtering, and automation around BTCUSDT on 15-minute data.