Brewed Insights: Forecasting Monthly Beer Production

Time series analysis is a powerful tool for understanding and forecasting temporal data. This project involves the analysis and prediction of monthly beer production using various time series forecasting techniques including SARIMA, LSTM(RNN) and Prophet models. The goal is to explore the data, understand its patterns, and apply forecasting models to predict future beer production.

About Data:

The dataset used in this project contains monthly beer production figures. It includes the following columns:

• Month: The month of production.
• Monthly beer production: The quantity of beer produced in that month.

Table of Contents:

1. Importing Libraries
2. Data Preprocessing
3. Data Exploration and Visualization
4. SARIMA Forecasting
5. LSTM (RNN) Forecasting
6. Prophet Forecasting
7. Model Comparison
8. Conclusion

Importing Libraries

• Pandas & NumPy for data manipulation and analysis.
• Matplotlib & Seaborn for data visualization, creating plots and charts.
• seasonal_decompose & SARIMAX from statsmodels to explore & forecast time series data.
• MinMaxScaler from sklearn to scale model input data.
• Sequential, Dense & LSTM from keras for building LSTM(RNN) model.
• Prophet from prophet library for time series forcasting.
• mean_squared_error & rmse from sklearn & statsmodels to evaluate model performance.

Data Preprocessing

• Loading the Data: The dataset is loaded into a Pandas DataFrame.
• Date Conversion: The Month column is converted to a datetime object and set as the index.
• Frequency Setting: The frequency of the datetime index is set to monthly ('MS').

Data Exploration and Visualization

• Basic Information: The structure and summary statistics of the data are explored.
• Seasonality: Plot the monthly beer production to visualize the overall trend and seasonal patterns in the data.
• Seasonal Decomposition: The seasonal_decompose function is used to decompose the time series into its trend, seasonal and residual components using the additive model to better understand the underlying patterns.

SARIMA Forecasting

• SARIMA Model: A Seasonal ARIMA (SARIMA) model to forecast future production based on past trends and seasonality.
• Model Fitting: A SARIMA model is fitted to the training data.
• Prediction: The model predicts beer production for the test period.
• Evaluation: The predictions are evaluated using RMSE and MSE metrics.

LSTM (RNN) Forecasting

• LSTM Model: A Long Short-Term Memory (LSTM) model, a type of Recurrent Neural Network (RNN), for time series forecasting.
• Data Scaling: The data is scaled using MinMaxScaler.
• Model Building: An LSTM model is built and trained on the scaled data.
• Prediction: The model predicts beer production for the test period.
• Evaluation: The predictions are evaluated using RMSE and MSE metrics.

Prophet Forecasting

• Prophet Model: Facebook's Prophet library, a versatile forecasting tool, to predict future production.
• Model Fitting: Trains the model on historical data and generates forecasts for upcoming months.
• Prediction: The model predicts beer production for the test period.
• Evaluation: The predictions are evaluated using RMSE and MSE metrics.

Model Comparison

• The performance of all three forecasting models (SARIMA, LSTM, Prophet) is compared using on RMSE and MSE metrics.
• Provides a table summarizing the error metrics for each model.
• Visualizes all predictions along with the actual production data for a comprehensive comparison.

Conclusion

This project demonstrates the power of time series analysis for uncovering patterns and predicting future trends in time-dependent data. By leveraging various forecasting techniques, valuable insights can be gained for informed decision-making within the beer production industry.