Hey everyone, I'm stuck on the feature engineering part of this problem. I'm not sure how to deal with all the different sales categories. I was thinking of maybe using the average sales value for each category and then doing some kind of normalization. But, honestly, I'm not sure. What do you guys think?
#🛍┊store-sales-time-series-forecasting
bright harness
prime roost
I haven't dug into this comp myself yet, but I'd be curious as to others' suggestions. Anyone have an idea?
polar plume
hey just started this competition but am pretty to this thing is there any team mate so we can work on this togather
prime roost
hey just started this competition but am pretty to this thing is there any team ...
Hey there, glad to have you here! Feel free to post in https://discord.com/channels/1101210829807956100/1130572338182762657 for teaming up.
craggy hound
Can I use ARIMA model to solve the problom,I think I can use it to predict time series?
iron salmon
Hopefully I'll be done with this comp in a couple hours, took a lot of research and time. And I am doing this in a very simple manner, and will be using XGboost.
glossy viper
https://www.kaggle.com/code/lorentzyeung/calendarfourier-deterministicprocess-fourier
CalendarFourier, DeterministicProcess, Fourier for time-series forecasting
Hello Kaggle Community! 👋
I'm sharing a Jupyter Notebook that aims to demystify some of the deterministic terms in Statsmodels' Time Series Analysis module, specifically CalendarFourier, DeterministicProcess, and Fourier. I am just uploading it to this competition (because I came across these terms in this particular competition) to share the knowledge around. I am not sure how to do this sharing right. Please let me know if this is the right way to do it, and also suggest me the right way to do it.
Why This Notebook?
Time series analysis is a crucial part of data science, especially in fields like finance, healthcare, and climate science. While working with time series data, we often encounter seasonality, trends, and other patterns that need to be modeled accurately for better forecasts. Statsmodels offers a variety of tools for this, but some of them can be a bit intimidating at first glance. Hope this notebook will help.
What's Inside?
CalendarFourier: Learn how to model complex seasonal patterns using Fourier series with real-world examples.
DeterministicProcess: Understand how to include deterministic components like constant terms, linear trends, and even quadratic trends in your time series model.
Fourier: A more general approach to Fourier series in time series, allowing you to specify the frequency and order directly.
Key Takeaways
Practical Python code examples for each deterministic term.
Visualizations to help you understand the impact of each term on the model.
Tips on how to choose the right parameters for your specific use-case.
I've tried to make this notebook as comprehensive as possible, but if there's anything unclear or you have any questions, feel free to ask. Your feedback is invaluable! Please give it a vote if you find it useful.🙂 🙂
knotty oar
Hello everyone.
A transactions.csv dataset is provided but not specified in the Files Description and Data Field Information section. This dataset has the same series range as the train.csv dataset. I'm having a little trouble figuring out why it was provided given that it's series only seems to apply to the training dataset. If you have more information with regards to how this series can be used or why it was provided, please let me know.
olive roost
What is the link of this competition?
knotty oar
What is the link of this competition?
teal veldt
Can I use ARIMA model to solve the problom,I think I can use it to predict time ...
technically you can but you'll have to train number_of_stores * number_of_families = 3354 ARIMA models !
merry crest
hey all I just started working on the competition, have a question about the feature #onpromotion in the train dataset what does this feature refer to?
alright! I just found what I am looking for in the file description section lol thanks anyway if u came across my question and thought of helping!
wild linden
Just finished my first generation of the model. There is a lot of room for improvement!
wild linden
What architecture of model have you guys used? One model for all the data, one model per store, one model per family, any kind of ensembled model or something different?
wild linden
I have achieved RMSLE=0.43149, the top RMSLE=37.620.
vocal holly
Hello everyone really great to join the discord! I'm a grad student taking a time series course this semester with teammates and we want to join the competition Profit Estimation- Time Series Forecasting (https://www.kaggle.com/competitions/profit-estimation-time-series-forecasting/overview). I joined the Discord to see if we could join so that we can do our project on this data set. We are citing everything as Kaggle. Does anyone know how one can get added to a competition that is closed? When I try to download the dataset I see the message "This is a limited-participation competition. Only invited users may participate." Any help would be appreciated! I believe this channel is the exact spot where I hope to receive help. Thanks!
edgy prairie
@vocal holly Unfortunately not all competitions on Kaggle allow for data access / late submissions after the competition is over. This varies from competition to competition because of different host rules and licenses.
You will probably need to use a different competition/dataset for your project.
trail cloud
Hi all, I've implemented SARIMA - got a RMSLE score of 0.55 😦 I have an alternative idea, and looking fo a teammate with good (better) coding skills to implement it together!
solemn bay
Hey all. I have been using deterministic process for this competition. Then, I use X=dp.in_sample()
My problem is, now I want to pull in the 'lag' for each sales nbr / family, before my data goes into LinearRegression(). However, I can't because the index for deterministic process is just the date.
Is there any way I can go about achieving this? Hope that makes sense
viral tendon
My code is taking more than 12minutes(stopped after 12 minutes manually) just to ask is it just me or you too.
final root
Anyone want to form a team in this competition?
covert bane
Hii