LSTM Price Prediction Guide

A guide to building LSTM (Long Short-Term Memory) neural networks for cryptocurrency price prediction. Covers the full pipeline from data collection to model deployment.

Pipeline Overview

Data Collection → Feature Engineering → Preprocessing → Model Training → Evaluation → Deployment
     │                    │                  │                │               │            │
 CoinGecko,         Indicators,         Normalize,       LSTM layers,    RMSE/MAE,    REST API
 Binance API      Sentiment, Volume     Sequence data    Dropout, Dense   Backtesting   or MCP

Data Collection

import pandas as pd
from cryptodatapy import DataRequest

# Fetch historical OHLCV data
dr = DataRequest(
    tickers=['SPA'],
    fields=['open', 'high', 'low', 'close', 'volume'],
    freq='1h',
    start_date='2023-01-01',
    source='binance'
)
data = dr.fetch()

Feature Engineering

Technical Indicators

import ta

def add_features(df):
    # Moving averages
    df['sma_7'] = ta.trend.sma_indicator(df['close'], window=7)
    df['sma_25'] = ta.trend.sma_indicator(df['close'], window=25)
    df['ema_12'] = ta.trend.ema_indicator(df['close'], window=12)
    
    # RSI
    df['rsi'] = ta.momentum.rsi(df['close'], window=14)
    
    # MACD
    macd = ta.trend.MACD(df['close'])
    df['macd'] = macd.macd()
    df['macd_signal'] = macd.macd_signal()
    
    # Bollinger Bands
    bb = ta.volatility.BollingerBands(df['close'])
    df['bb_upper'] = bb.bollinger_hband()
    df['bb_lower'] = bb.bollinger_lband()
    
    # Volume indicators
    df['volume_sma'] = df['volume'].rolling(window=20).mean()
    df['volume_ratio'] = df['volume'] / df['volume_sma']
    
    # Returns
    df['returns'] = df['close'].pct_change()
    df['log_returns'] = np.log(df['close'] / df['close'].shift(1))
    
    return df.dropna()

Feature List

Feature	Type	Description
SMA (7, 25, 50)	Trend	Simple moving averages
EMA (12, 26)	Trend	Exponential moving averages
RSI (14)	Momentum	Relative strength index
MACD	Momentum	Moving average convergence
Bollinger Bands	Volatility	Price channels
Volume Ratio	Volume	Relative volume
Returns	Price	Percentage returns
ATR	Volatility	Average true range

Data Preprocessing

from sklearn.preprocessing import MinMaxScaler
import numpy as np

# Scale features to [0, 1]
scaler = MinMaxScaler()
scaled_data = scaler.fit_transform(features)

# Create sequences for LSTM
def create_sequences(data, seq_length=60):
    X, y = [], []
    for i in range(seq_length, len(data)):
        X.append(data[i - seq_length:i])
        y.append(data[i, 0])  # Predict close price
    return np.array(X), np.array(y)

X, y = create_sequences(scaled_data, seq_length=60)

# Train/validation/test split (70/15/15)
train_size = int(len(X) * 0.7)
val_size = int(len(X) * 0.15)

X_train, y_train = X[:train_size], y[:train_size]
X_val, y_val = X[train_size:train_size+val_size], y[train_size:train_size+val_size]
X_test, y_test = X[train_size+val_size:], y[train_size+val_size:]

LSTM Model

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dense, Dropout
from tensorflow.keras.callbacks import EarlyStopping

model = Sequential([
    LSTM(128, return_sequences=True, input_shape=(X_train.shape[1], X_train.shape[2])),
    Dropout(0.2),
    LSTM(64, return_sequences=True),
    Dropout(0.2),
    LSTM(32, return_sequences=False),
    Dropout(0.2),
    Dense(16, activation='relu'),
    Dense(1)  # Price prediction
])

model.compile(optimizer='adam', loss='mse', metrics=['mae'])

history = model.fit(
    X_train, y_train,
    validation_data=(X_val, y_val),
    epochs=100,
    batch_size=32,
    callbacks=[EarlyStopping(patience=10, restore_best_weights=True)]
)

Evaluation

from sklearn.metrics import mean_squared_error, mean_absolute_error

predictions = model.predict(X_test)

# Inverse transform to get actual prices
predictions_inv = scaler.inverse_transform(...)
actual_inv = scaler.inverse_transform(...)

rmse = np.sqrt(mean_squared_error(actual_inv, predictions_inv))
mae = mean_absolute_error(actual_inv, predictions_inv)

print(f"RMSE: ${rmse:.4f}")
print(f"MAE: ${mae:.4f}")

Metrics to Watch

Metric	Good	Acceptable
RMSE	< 2% of price	< 5%
MAE	< 1.5% of price	< 4%
Direction Accuracy	> 60%	> 55%
Sharpe Ratio (backtest)	> 1.5	> 1.0

Disclaimer

⚠️ LSTM predictions are for educational purposes only. Crypto markets are highly volatile and unpredictable. Never use model predictions as the sole basis for trading decisions. Past performance does not guarantee future results.

Links

• GitHub: https://github.com/nirholas/LSTM-price-prediction
• TensorFlow: https://www.tensorflow.org
• ta-lib: https://ta-lib.org
• Sperax: https://app.sperax.io