src/model_training.py

import pandas as pd
import numpy as np
from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score, classification_report
import joblib
import os
import logging



def load_features_target(horizon):
    print(f"{horizon} - HORIZON")
    X = pd.read_csv(f"data/features_{horizon}m.csv", index_col=0)
    y = pd.read_csv(f"data/target_{horizon}m.csv", index_col=0).squeeze()
    print(f"Tô aqui: {X, y}")
    return X, y


def temporal_train_split(X, y, test_size=0.2):
    n = len(X)

    if n == 0:
        return None, None, None, None

    split_idx = int(n * (1 - test_size))

    X_train = X.iloc[:split_idx]
    X_test = X.iloc[split_idx:]

    y_train = y.iloc[:split_idx]
    y_test = y.iloc[split_idx:]

    return X_train, X_test, y_train, y_test


def train_model(horizon, model_type='random_forest', save_model=True):

    X, y = load_features_target(horizon)

    print(f"\n========== Horizonte {horizon} ==========")
    print("Shape X:", X.shape)
    print("Shape y:", y.shape)

    if len(X) == 0:
        print("❌ Dataset vazio. Pulando treino.")
        return None

    X_train, X_test, y_train, y_test = temporal_train_split(X, y)

    if X_train is None or len(X_train) == 0:
        print("❌ Sem dados de treino.")
        return None

    if model_type == 'logistic':
        model = LogisticRegression(max_iter=1000, random_state=42)
    else:
        model = RandomForestClassifier(
            n_estimators=100,
            random_state=42
        )

    print("Treinando modelo...")

    model.fit(X_train, y_train)

    if len(X_test) == 0:
        print("⚠️ Sem dados de teste. Modelo treinado mas sem avaliação.")
        return model

    y_pred = model.predict(X_test)

    acc = accuracy_score(y_test, y_pred)

    print(f"Acurácia: {acc:.4f}")

    report = classification_report(
        y_test,
        y_pred,
        target_names=['Queda', 'Sobe']
    )

    print(report)

    if save_model:
        os.makedirs("models", exist_ok=True)

        path = f"models/{model_type}_{horizon}m.pkl"

        joblib.dump(model, path)

        print("Modelo salvo em:", path)

    return model


def train_all_horizons():

    horizons = [10, 24, 60, 120]

    print("\n===== Random Forest =====")

    for h in horizons:
        train_model(h, 'random_forest')

    print("\n===== Logistic Regression =====")

    for h in horizons:
        train_model(h, 'logistic')