lstm-petr4-stock-prediction

import torch
import joblib
import json
import numpy as np

# Carregar artefatos
config = json.load(open("config.json"))
scaler = joblib.load("scaler.joblib")

# Reconstruir modelo
from model import LSTMModel

model = LSTMModel(
    input_size=config["model"]["input_size"],
    hidden_size=config["model"]["hidden_size"],
    num_layers=config["model"]["num_layers"],
    dropout=config["model"]["dropout"]
)
model.load_state_dict(torch.load("lstm_model.pth", map_location="cpu", weights_only=True))
model.eval()

# Fazer previsao (com 60 precos de fechamento)
prices = np.array([...]).reshape(-1, 1)  # 60 precos
prices_scaled = scaler.transform(prices)
input_tensor = torch.FloatTensor(prices_scaled.reshape(1, 60, 1))

with torch.no_grad():
    prediction = model(input_tensor).item()

predicted_price = scaler.inverse_transform([[prediction]])[0][0]
print(f"Preco previsto: R$ {predicted_price:.2f}")