Feature Engineering for Time Series Forecasting - Kishan Manani

preview_player
Показать описание
In this podcast episode, we talked with Kishan Manani about feature engineering for time series forecasting.

0:00 Introduction and Welcome
2:16 Speaker Introduction
2:54 Topic Introduction: Feature Engineering for Time Series Forecasting
4:23 Motivating Example: M5 Forecasting Competition
6:25 Machine Learning for Time Series Forecasting
8:50 Direct Forecasting vs. Recursive Forecasting
10:50 Creating Lag Features
11:45 Handling Exogenous Variables
15:55 Static Features
18:00 Time Series Cross Validation
20:00 Key Differences in Machine Learning Workflow
21:35 Feature Engineering Overview
23:00 Lag Features and Correlation Methods
29:20 Window Features
32:25 Static Features and Encoding
37:25 Avoiding Data Leakage
39:30 Useful Libraries and Tools
40:30 Example with Darts Library
45:00 Conclusions and Q&A

🔗 USEFUL LINKS

🎙 ABOUT THE PODCAST
At DataTalksClub, we organize live podcasts that feature a diverse range of guests from the data field. Each podcast is a free-form conversation guided by a prepared set of questions, designed to learn about the guests’ career trajectories, life experiences, and practical advice. These insightful discussions draw on the expertise of data practitioners from various backgrounds.

We stream the podcasts on YouTube, where each session is also recorded and published on our channel, complete with timestamps, a transcript, and important links.

📚Check our free online courses

👋🏼 GET IN TOUCH

Рекомендации по теме
Комментарии
Автор

It was a great talk about data. Thank you so much. I hope you can share similar talks on the future as well

iftikhar
Автор

Thanks for this informative video! 👏👏👏

ivanliu
Автор

good to know :D I am working on RUL estimating and prognosis using time series data.

MinhVu-ymtk
Автор

Great presentation. To clarify, is overfitting always an issue? I'm assuming it always is. In the scenario where you compute the window values, ensuring you're only using the available data... there will be no leakage at a row-level. But when you consider all training values.. for example at Time = 1 vs Time = 8, the relationships being built by the Forecasting algorithm when predicting Time = 1 will still use Time = 8 values.

jacobschultz
Автор

What tools are folks using to expose/extract/generate features? Tsfresh? getML? I work in Java for my ML tasks but will happily integreate Python or C/C++ based tools into the pipeline.

I'm not a statistics guy so I can't write these feature generation algos myself.

bueocean
Автор

In target encoding section, when product ID is encoded dynamically, how will the model distinguish between the data points belonging to same time series or different time series?

gurjinderkaur
Автор

It is great but something is wrong with time_col in definition of the procedure. It seems to work if that column is an index and not mentioned in a function call.

piotrbjastrzebski
Автор

Hi, does anyone know how to implement the recursive forecasting that he did in Darts using sktime. I couldn't really find an intuitive explanation online.

AhmedThahir
Автор

For two different time series, does it make sense to build two separate models instead of having the targets of both the series in the single model (as shown at 24:40)?

anoubhav
Автор

can you explain something about stock prediction?

pranavkhatri
Автор

What is the name/link of the “chunky” review paper you mentioned at the end of the presentation?

RDarrylR
Автор

can someone help me to deal with categorical features for forecasting time series in ML

mamyrak