We recently came across a python opensource package (https://pypi.org/project/th2analytics/) that is truly revolutionary in the time series forecasting space.
In our benchmarks, it consistently outperformed Nixtla’s libraries and even AWS Chronos, both in terms of accuracy and scalability. The framework feels like a real step change — not just an incremental improvement — making it one of the most promising tools for production-grade forecasting we’ve tested so far.
Anis_FA•4mo ago
Follow-up: I should also add that what really sets this package apart is the balance between accuracy and efficiency. In our tests, it not only delivered lower MAPE and sMAPE than Nixtla and Chronos, but also trained much faster. That combination makes it extremely compelling for real-world, production-level forecasting.
Anis_FA•4mo ago
In our benchmarks, it consistently outperformed Nixtla’s libraries and even AWS Chronos, both in terms of accuracy and scalability. The framework feels like a real step change — not just an incremental improvement — making it one of the most promising tools for production-grade forecasting we’ve tested so far.
Anis_FA•4mo ago