Giannis Tolios’ Post

𝗦𝗶𝗺𝗽𝗹𝗶𝗳𝘆 𝗧𝗶𝗺𝗲 𝗦𝗲𝗿𝗶𝗲𝘀 𝗙𝗼𝗿𝗲𝗰𝗮𝘀𝘁𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗗𝗮𝗿𝘁𝘀! The Darts library has simplified time series analysis and forecasting with Python. Darts supports various forecasting approaches, ranging from statistical models like ARIMA, to novel methods based on deep learning. Darts also supports advanced techniques like explainable forecasting, conformal prediction and anomaly detection. Therefore, Darts has been established as one of the best libraries for time series tasks, making it extremely useful to data scientists and researchers. Visit the link below for more information and follow me for regular data science content! 𝗗𝗮𝗿𝘁𝘀 𝗹𝗶𝗯𝗿𝗮𝗿𝘆 𝘄𝗲𝗯𝘀𝗶𝘁𝗲: https://lnkd.in/dEQepm3D 𝗟𝗲𝗮𝗿𝗻 𝗠𝗟 𝗮𝗻𝗱 𝗙𝗼𝗿𝗲𝗰𝗮𝘀𝘁𝗶𝗻𝗴: https://lnkd.in/dyByK4F #datascience #python #machinelearning #deeplearning

  • No alternative text description for this image

Does Darts support FB prophet and Kats?

darts is such a good recommendation!! thank you for bringing this up. It's by far one of my favorite libraries for its support of both univariate and multivariate time series and models

The gap between demand forecasting accuracy and actual inventory optimization outcomes is where most implementations stall. Even a model with 95% forecast accuracy can produce poor inventory decisions if the cost function is wrong. Symmetric loss functions like RMSE penalize over-prediction and under-prediction equally, but in supply chains the costs are asymmetric. Stockout costs and holding costs differ by orders of magnitude depending on the product category. The more impactful research direction is combining probabilistic forecasting with decision-theoretic optimization. Rather than predicting a single demand value, generating full probability distributions allows the inventory optimizer to make cost-sensitive decisions under uncertainty. Have you explored how well your AI-driven forecasts calibrate at the tails of the distribution? Most neural forecasters perform well around the median but badly underestimate the probability of extreme demand spikes, which is precisely where inventory decisions matter most.

Really handy library for evolving behaviours Ankita Saini , have a look

See more comments

To view or add a comment, sign in

Explore content categories