This project demonstrates the use of AutoGluon: Time Series Forecasting for forecasting iron ore prices.

Evaluation and Explanation

There might be some confusion between the MAE value calculated in the model and the MAE reported by AutoGluon. AutoGluon calculates the MAE during model training using the validation set. However, the MAE value you calculate in the model represents the error between the predicted and actual values for the entire dataset. AutoGluon Validation MAE During model training, AutoGluon splits the dataset into training and validation sets. The validation set is used to evaluate the model's performance during training. The following MAE values are reported for different models:

• ETS Model: Validation MAE = 93.33

• Theta Model: Validation MAE = 88.39

• DeepAR Model: Validation MAE = 28.41

• Weighted Ensemble Model: Validation MAE = 15.41 (Best model)

These values show the model's performance on the validation set. The Weighted Ensemble Model, which combines the predictions from DeepAR and Theta models, performed the best with the lowest validation MAE.

This project demonstrates the use of AutoGluon: Time Series Forecasting for forecasting iron ore prices. The workflow includes data preprocessing, feature engineering, model training, evaluation, and visualization. The results show the model's predictions and the calculated Mean Absolute Error (MAE), providing insights into the model's performance. This is a univariate time series forecasting project, focusing on the prediction of iron ore prices based on historical price data. Various models including ETS, Theta, and DeepAR (Neural Network) were utilized, and a weighted ensemble model was used to select the best-performing model.