# Let's start by turning off warnings for a cleaner experience
import warnings
warnings.filterwarnings("ignore")

# Download the shampoo dataset and check the contents
import pandas as pd
url = "https://raw.githubusercontent.com/jbrownlee/Datasets/master/shampoo.csv"
df = pd.read_csv(url)
print("Data shape:", df.shape)
df.head()

# Make a simple line plot to visualize the sales over time
import matplotlib.pyplot as plt
plt.plot(df["Month"], df["Sales"])
plt.xlabel("Month")
plt.ylabel("Sales")
plt.title("Monthly Shampoo Sales")
plt.xticks(rotation=45)
plt.show()

# Create 'lag' features for machine learning
df["Sales_prev1"] = df["Sales"].shift(1)
df["Sales_prev2"] = df["Sales"].shift(2)
df = df.dropna()
df.head()

# Split the data into train and test sets
train = df.iloc[:-6]
test = df.iloc[-6:]
print("Train shape:", train.shape)
print("Test shape:", test.shape)

# Fit Random Forest to the train set
from sklearn.ensemble import RandomForestRegressor
X_train = train[["Sales_prev1", "Sales_prev2"]]
y_train = train["Sales"]
rf = RandomForestRegressor(random_state=42)
rf.fit(X_train, y_train)

# Predict sales for the test set and check results
X_test = test[["Sales_prev1", "Sales_prev2"]]
y_test = test["Sales"]
y_pred_rf = rf.predict(X_test)
print("Random Forest predictions:", y_pred_rf)
print("True Sales:", y_test.values)

# Plot the predicted vs. actual sales
plt.figure(figsize=(8,4))
plt.plot(test["Month"], y_test, label="Actual")
plt.plot(test["Month"], y_pred_rf, label="Random Forest Predicted")
plt.xlabel("Month")
plt.ylabel("Sales")
plt.title("Random Forest: Actual vs Predicted Sales")
plt.legend()
plt.xticks(rotation=45)
plt.show()

# If running locally, you may need to install xgboost first
# Uncomment and run the next line if needed:
# !pip install xgboost

# Train XGBoost on our sales data
from xgboost import XGBRegressor
xgb = XGBRegressor(random_state=42)
xgb.fit(X_train, y_train)
y_pred_xgb = xgb.predict(X_test)

# Compare XGBoost predictions to actual sales
print("XGBoost predictions:", y_pred_xgb)
print("True Sales:", y_test.values)

# Plot XGBoost and Random Forest together
plt.figure(figsize=(8,4))
plt.plot(test["Month"], y_test, label="Actual")
plt.plot(test["Month"], y_pred_rf, label="Random Forest")
plt.plot(test["Month"], y_pred_xgb, label="XGBoost")
plt.xlabel("Month")
plt.ylabel("Sales")
plt.title("Comparison: Actual vs Random Forest vs XGBoost")
plt.legend()
plt.xticks(rotation=45)
plt.show()

# Evaluate accuracy with MAE
from sklearn.metrics import mean_absolute_error
mae_rf = mean_absolute_error(y_test, y_pred_rf)
mae_xgb = mean_absolute_error(y_test, y_pred_xgb)
print("Random Forest MAE:", mae_rf)
print("XGBoost MAE:", mae_xgb)

# Predict the next month's sales using the final test lags
last_row = test.iloc[-1]
future_lags = [[last_row["Sales_prev1"], last_row["Sales_prev2"]]]
rf_next = rf.predict(future_lags)[0]
xgb_next = xgb.predict(future_lags)[0]
print("Random Forest predicts next month's sales as:", round(rf_next, 2))
print("XGBoost predicts next month's sales as:", round(xgb_next, 2))

# Try getting a feature from user input
print("Enter a value for sales two months ago:")
lag2 = float(input())
print("Enter a value for sales one month ago:")
lag1 = float(input())
guess = rf.predict([[lag1, lag2]])[0]
print("Random Forest would predict:", round(guess, 2))