Method

Time-series forecasting

Predicting future values of a series incl. demand and metrics.

45 cases from 31 companies apply Time-series forecasting. 2 were checked in an online A/B test; 18 report results in production. Includes variants: ARIMA / VAR, Prophet.

Typical tasks: Forecast demand and load, Detect anomalies, Optimize decisions, Build a data platform and 10 more. Companies: Yandex Lavka, Airbnb, Lyft, Walmart, Uber, Avito and 25 more. Often combined with: Prophet, Anomaly detection, ARIMA / VAR, Gradient boosting, Linear regression and 82 more.

How we knew COVID was over (and what our models had to unlearn)travel2026-08
Online Learning for Cost-Efficient LLM Routingfintech2026-07
Forecasting at Databricks: One Framework Behind Consumption and Infra Costsaas-b2b2026-06
When history fails you, borrow from geographytravel2026-06
ML Ad Forecaster: Estimating Results Before Launchclassifieds2026-05
Method of Automated Construction of Production Digital Twins Using LLM Agentsmanufacturing2026-05
What COVID did to our forecasting models (and what we built to handle the next shock)travel2026-03
Pricing in Yandex Lavkafood-delivery2025-11
Recommendation and Optimization. How to Recommend Cashback to Customers and Not Go Broke?banking2025-05
Virtual Sensors in Industrymanufacturing2025-05
Optimizing Regular Prices: Combining Experimental Approaches, ML and Economic Theoryretail2025-05
Building State-of-the-Art Tracking with ReIDgeneral2025-05
A Short-Term Anomaly Forecasting Model for Univariate Time Series and Its Use in the Sfera.IAS Productgeneral2025-05
Forecasting Model Architecture in Yandex Lavkafood-delivery2025-05
Real-Time Spatial Temporal Forecasting @ Lyftmobility2025-05
Autotuning: A comprehensive approach to Time Series Forecastingretail2025-03
Decoding Causal Incrementality in E-Commerce: Leveraging Bayesian Structural Time Series Model with a Real-World Examplee-commerce2025-01
Part 2: A Survey of Analytics Engineering Work at Netflixmedia-streaming2025-01
How We Cut Forecasting Time from Two Weeks to Five Minutesclassifieds2024-05
Short-Term Order Forecasting for Creating Courier Slots at Dark Storesfood-delivery2024-05
Machine Learning for Flotation Optimizationmanufacturing2024-05
Foundation Models: Is There Life in Time Series?general2024-05
Forecast and chill: Zero-Shot Foundation Models for Time Seriesgeneral2024-05
Anomaly Detection: Anomaly Scoring for Preventive Detection of Information System Failuresbanking2024-05
TSAD: A Python Library for Solving ML Problems in Industrymanufacturing2024-05
Economics of the Data Officegeneral2024-05
Real-world Insights: Anomaly Detection in Internet Traffictravel2024-02
Real-World Forecasting with Deep Learning: How We Do It at Wixsaas-b2b2024
The Curse of Data, Machine Sorcery and Dark Predictionmanufacturing2023-10
Order Forecasting on a Map Using a Single Time Seriesgeneral2023-10
Using Synthetic Search Data for Flights Price Forecastingtravel2023-10
How DoorDash Built an Ensemble Learning Model for Time Series Forecastingfood-delivery2023-06
ML for Forecasting and Analyzing Passenger Transport Behavior—mobility2023-05
Modeling the Behavior of an Oil Spillmanufacturing2023-05
Preprocessing and Anomaly Detection in Time Seriesmanufacturing2023-05
Building a large scale unsupervised model anomaly detection system — Part 2mobility2023-04
Demand and ETR Forecasting at Airportsmobility2023-03
D3: An Automated System to Detect Data Driftsmobility2023-02
Using mobile sensor data to encourage safer drivingmobility2022-10
Causal Forecasting at Lyft (Part 1)mobility2022-06
Leveraging Causal Inference to Generate Accurate Forecastsfood-delivery2022-06
How We Built Infrastructure to Run User Forecasts at Spotifymedia-streaming2022-06
Fitting a univariate time series model to a cluster of seriesretail2022-03
Project RADAR: Intelligent Early Fraud Detection System with Humans in the Loopmobility2022-02
Building for Balancemarketplace2022-01