create a new dynamodb for all the results of the current best model (xgboost), and present them in the streamlit interface

make something like a gru to make long term predictions

show stats like:
-top 5 most profitable companies (given by best eps for the last quarter)
-show another column for the histogram indicating the future estimated eps
-top 5 most surprising companies (given by biggest surprise percentage)

recreate keys and assign them to all services that need it

change connectivity of efs to support only calls from ec2 and lambda

add graphs to the readme page and update the about page of the app.
Also change the aws diagram

sample from alpha vantage

In the stock reccomendations, we would like to see as pop ups stocks that are trending at news and also have an anomalous volume of trading

ec2 instances are overkill. move it to lambda functions with a preloaded venv

since dynamodb costs start to pile up, investigate other options such as moving to S3 as storage and saving dataframes in parquet.
check whether you could get sql like access from it then

also lower the ec2 time to 2 hours

upload all files to a new dynamoDB table called models

Once there are new graphs, Update the readme file with:

1. New Graphs (and maybe the streamlit page example and api example) - Avihai
2. Documentation of what the repo does in general - Dudu

only if possible to collect as usual

1. Create a public api to gather data from the dynamoDB's you have in the cloud
2. Connect streamlit functionality to pull the data from the new api

Once theres a new and improved model for analysis, upload the model in inference mode to production in a dedicated ec2 container.
also build an easy pipeline for uploading retrained models/ new models

page width is currently only set when going through the main page. add an if statement that checks if its set to wide and if it doesnt, sets it to wide

Create some data regarding which news outlets give the best prediction regarding stock prices and post visualizations to streamlit api

create a trading bot that will test what will be the profit if we use the model for trading (in a seperate .py file)

dynamodb is a nosql database, which is unfit for storing relational data such as stock data.
move tables to rds, and store the saved models in dynamodb

show a histogram of earnings in the stock data page including future estimated earnings

learn SQL and move all databases

In the streamlit web app api, present actual database data

since dynamodb doesnt handle duplicates, add a container that scans all databases periodically and deletes the duplicates

this line throws an exception (probably went unnoticed because it was only a print there):
if "time" not in kwargs: raise Exception(f"Invalid use of the cache_dec decorator. {func} must have a time variable")

in dataloader.py

Create an EC2 container in aws that scrapes the stock prices and updates the dynamoDB table accordingly.
Connect the container to a lambda function and an event bridge watcher so it will be activated once a day.

show something like a pie chart of the companies balance sheets from the database

update the stock recommendation page with real time predictions for each requested stock

improve prediction of stock prices by:

1. hyper param search on XGBoost
2. adding monthly earnings to the dataset
3. using DNN such as LSTM, GRU, Transformers (last resort, probably sucks) etc...

add a data base for earnings that is scraped and built once a month for all stocks.
the container will also concatenate the data to the stock price database

in the news sentiment page, let the user see links to the most relevant articles for each stock

now that the app is connected to the database. create real plots and graphs of the data so well have a presentable first rough draft 🥇