Forecasting using N Hits

Max Mergenthaler Canseco and Federico Garza Ramírez - Forecasting using N Hits In this recording Max Mergenthaler Canseco and Federico Garza Ramírez are describing the challenges of long-horizon forecasting, such as the volatility of predictions and computational complexity, and introduces a new model called N-HiTS that addresses these challenges by incorporating novel hierarchical interpolation and multi-rate data sampling techniques. The hierarchical interpolation technique can efficiently approximate long horizons in the presence of smoothness, and extensive experiments on large-scale datasets demonstrate that N-HiTS outperforms state-of-the-art methods, with an average accuracy improvement of 16% and a 50 times reduction in computation time. The code for N-HiTS is available on GitHub.