I'm working on a time-series problem, and after running ExponentialSmoothing with different configurations I got the best two models shown in the screenshot below where first one is based on MAE performance while the other one is based on MSE. I have tunned the following hyper-parameters using nested for-loops: alpha, beta, gamma, seasonal_periods, trend, seasonal, use_boxcox.
- Does the model output look good or it should be enhanced further? if so, how can I enhance further?
- is there a better way to fine-tune a model, unforetenately cannot use RandomSearchCV
- any other algorithm should I look into?