Actually, with a different approach. A reckoning on the anomaly. Data requirement vs. normal PDMs. He sees that there is no data, but he tries to see if the system can train itself without any data. An anomaly arose in the first place. (Two model gates. Independent variable and multivariate.) Next, we'll see if these anomalies warrant a state of care. The reason for this is that, especially in our researches, seasonal conditions, tool life, engines, etc. We have found that it is effective in Seasonal conditions are very different in Turkey, especially in some cities. (An average of +35 degrees in summer, -15 degrees in winter.) If you have any suggestions, I'm open 🙂