The Fifth Elephant 2023 Monsoon

Link to presentation:

https://drive.google.com/file/d/1hH90FCWxRFv0IQcoBqX2yRM1DcTzDBQr/view?usp=sharing

Abstract:

This talk addresses time-series modeling for demand forecasting, covering a brief history of various types of forecasting models and showcasing where each class of methods is applicable. The talk also covers a case study with real-world data from Sortly (an inventory management SAAS company), highlighting the challenges faced with diverse consumption patterns and anomalies. Specifically, the results of forecasting with SARIMAX, Random Forests and a Transformer based model are discussed for this case-study. Additionally, the talk shares some behind the scenes prompt engineering stories for this case-study - including some detailed prompts, what worked and what did not.

What will you get out of this talk:

By attending this talk, attendees will gain insights into various types of time-series models and their effectiveness in demand forecasting for various types of data.

Talk Outline:

Part 1: A brief History of Demand Forecasting:
Various models for demand forecasting
Types of Models
Decomposable Time Series Models - Ex: Arima, Sarimax, Prophet
Classical ML and Ensemble based approaches - Ex: Random Forest, LightGBM
Deep learning based models - Example: LSTMs, GRU based model, Transformers
Where do each of these models make sense to use

Part 2: Demand Forecasting for Sortly
Understanding Sortly Data
Challenges - Anomalies, diverse workflows
Comparison of models for select data

Part 3: Prompt Engineering Stories
Where all can one use prompt engineering
Examples of detailed prompts
Examples of problems encountered with prompts
Prompting Lessons