Suggestion: Increase the display of the number of papers. about time-series-forecasting-and-deep-learning HOT 3 OPEN

Thanks for the suggestions and they are very useful!

1. I've been thinking about changing the structure and display style of the project as more papers are included and the whole list grows too long, in the future I will add some statistics like total number of paper, number of papers by year, etc. Also, I'm considering to add some labels to every paper or library to show which time series task it works on (forecasting, anomaly detection, feature generation...) It will take some time and I plan to finish it in the next several weeks.
2. I created this project with a focus on time series forecasting, but my interest is not limited to this. I want to make this project a resource collection that includes various time series tasks in a broad sense. Papers and resources about any time series related tasks are welcome and definitely helpful, including anomaly detection.
3. Let me know if you have any other questions or suggestions, I'd be happy to make the project better. @943fansi

from time-series-forecasting-and-deep-learning.

@DaoSword Dividing tasks and adding tags are good ideas. It can be borrowed from the structure of the review paper. Keywords in papers can be extracted as tags. The existing division method is not deep enough, and the depth of the structure does not exceed 3. There are usually multiple partitions, and occasionally a secondary partition. I hope you can build a tag tree that can be explored top-down, bottom-up, or from a node in the middle.

There are two commonly used sorting methods: one is sorted by time, and the other is sorted by popularity. I think sorting by time is good, and sorting by popularity needs to obtain information and update it. Maybe try to select Top k papers. Some websites support subscriptions and push new papers. When I searched on GitHub, I found many people's paper records. Keep the existing chronological order :).

The addition of tags should be able to quickly and accurately locate papers. If you know the title of the paper, just use the search engine directly. Tags and essay titles should not overlap. Some repositories are categorized by the conference in which they were published, which may be more convenient for organizing papers. But personally I think the reading experience is not good. Why not go directly to the official website of the conference to view the papers for retrieval?

from time-series-forecasting-and-deep-learning.

@943fansi These are pretty good suggestions. I agree with the idea that sorting by time is better, and I will keep the existing chronological order :). For tags and task division, I will try to build a tag tree or design a category system that is resonable and clear while keep things simple.

from time-series-forecasting-and-deep-learning.