Operational Analytics Project

Develop by: Negri Andrea, Spadoni Alberto

Dataset

https://www.kaggle.com/cityofLA/los-angeles-museum-visitors (Avila Adobe Museum)

Goal

Build a predictive model to forecast the amount of future museum's visitors

Highlights

• Trend identification
• Season identification (with seasonal coefficients)
• Model's type identification (additive or multiplicative)
• Combination of trend and seasonal coefficients to find a first simple "model" used to forecast 24 time periods
• SARIMA model used to forecast 24 time periods
• MLP model used to forecast 24 time periods
• LSTM model used to forecast 24 time periods
• Random Forest model used to forecast 24 time periods
• Diebold Mariano test to compare neural models (MLP-LSTM, MLP-RF, LSTM-RF)