Hhjff
Perfect, now I see the full details. Andrew says: He’s testing TensorFlow + XGBoost models for ASW 5d prediction. His results are decent but not great (RMSE ~1.5, MAPE 4.5). He's facing overfitting issues. He compliments your Random Forest approach (non-ML ensemble) saying it looks excellent. He also likes your regime shift model ("Very cool"). Thanks to this, here’s an even better, polished reply you can send: Subject: RE: Key Information: Regime Change Detected for ASW5Y Hi Andrew, Thank you for your email, and I’m really glad to hear you’re getting to spend such precious time with your daughter — enjoy every moment! It’s great to hear that you’re experimenting with TensorFlow and XGBoost for the ASW5Y prediction. Your results are actually very solid given the complexity of the signal — although I completely understand the challenge with overfitting. It’s definitely a tricky balance when using more flexible models, especially on financial time series ...