10-K reports discussion

[stancsz]

This post was supposed to be a welcome post, you keeners 👍. Please see below for the 10-K reports convo:
🚀🌕

[markomijovic]

@rmsouza01 @mklasby Stan suggested I ping you guys here regarding SEC filings API. https://sec-api.io/ is one of the most popular and best 10-K and other SEC filings APIs. Pretty much returns SEC filings for all US companies. 100 free calls and after that costs ~50USD/month

[rmsouza01]

@markomijovic: I am not very familiar with the 10-K reports. I guess some of the features you are planning would come out of the report?

Do you know if the reports are structured (i.e., easy to mine for information)? If not, you are thinking about mining the reports using natural language processing to extract the information?

I would suggest starting with the free calls and see if a subscription is worth it.

[rmsouza01]

Sorry @stancsz, we should have created a new discussion to talk about the 10-k reports.

[markomijovic]

@rmsouza01 Hey, I've read a few 10-K reports. That API returns all numeric data as a json document so no need to mine information. The rest of the text is objective on purpose so no one can interpret it as positive/negative.

My idea was that the information you'd find in SEC filings like sales by product/by region, total debt, debt rates / or debt issued by year as a % of operating income / shares outstanding vs shares issues in a year etc etc. would be useful for other ML models you have. I think having more labelled data like that would help make better models although I'm not familiar with what other work you've done yet.

In addition, I could use discounted cash flow which essentially considers time value of money to get a single net worth of a company in today's dollars. Usually though this would be very different from the market value since market prices future growth, however by having present and market value, we can find the priced in % growth in X # of years. So then say market is pricing in higher % growth but SEC filing shows China sales are on-decline, well probably not a good sentiment in terms of hitting that %.

I'll def get started and explore this more in a week or two once capstone is done.

[stancsz]

Thanks Roberto & thanks Marko for sharing this! I think this is definitely an interesting data source / knowledge source. This should fall under our pillar of big data / data mining.

I suggest we do some EDA & meeting sessions to discuss the feasibility of integrating this data source. Until we have an actual product volunteer to keep track of the roadmaps, let's use discussion for this purpose.

@markomijovic since you brought this initiative up, would you like to drive this module of this project? owning this bigdata/unstructured project means you keep track of the progress & share your research, kinda like being your own product. After we scoped out what to do, some engineering members can tackle this either using spark or other tools. FYI @mklasby @Karenzhang7717

[markomijovic]

@stancsz Yea definitely, I can drive this module

[stancsz]

Hi @markomijovic thanks for the taking initiative, to help you get started, I've put together a starter project for you
https://github.com/deeptendies/mining-sec-edgar

for the most interesting data: financial reports, the reports is already avaliable at:

parsed_filings/0000950123-10-030164/0.Financial_Report.xls

parsed_filings/0001326380-18-000033/0.Financial_Report.xlsx

feel free to write codes to load & combine them
suggestions:
https://pandas.pydata.org/docs/reference/api/pandas.read_excel.html

have fun & feel free reach out to myself and @mklasby for engineering assistance, & we can work on this for future roadmap to incorporate this extraneous as additional source of knowledge