import tensorflow as tf
import pandas as pd
from sklearn.preprocessing import LabelEncoder, StandardScaler
from sklearn.model_selection import train_test_split
import json
import mlflow

mlflow.set_tracking_uri("http://localhost:5000")  # Ustawienie adresu MLflow Tracking Server

df = pd.read_csv('OrangeQualityData.csv')

encoder = LabelEncoder()
df["Color"] = encoder.fit_transform(df["Color"])
df["Variety"] = encoder.fit_transform(df["Variety"])
df["Blemishes"] = df["Blemishes (Y/N)"].apply(lambda x: 1 if x.startswith("Y") else 0)

df.drop(columns=["Blemishes (Y/N)"], inplace=True)

X = df.drop(columns=["Quality (1-5)"])
y = df["Quality (1-5)"]

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)

model = tf.keras.Sequential([
    tf.keras.layers.Dense(64, activation='relu', input_shape=(X_train_scaled.shape[1],)),
    tf.keras.layers.Dense(32, activation='relu'),
    tf.keras.layers.Dense(1)
])

model.compile(optimizer='sgd', loss='mse')

with mlflow.start_run():
    mlflow.log_param("optimizer", 'sgd')
    mlflow.log_param("loss_function", 'mse')
    mlflow.log_param("epochs", 100)

    history = model.fit(X_train_scaled, y_train, epochs=100, verbose=0, validation_data=(X_test_scaled, y_test))

    for key, value in history.history.items():
        mlflow.log_metric(key, value[-1])  # Logujemy ostatnią wartość metryki

    model.save('orange_quality_model_tf.h5')

    predictions = model.predict(X_test_scaled)

    with open('predictions_tf.json', 'w') as f:
        json.dump(predictions.tolist(), f, indent=4)