Neural Network Artificial Intelligence

This program consists of using financial information, that is custom to requirements; it uses 10 years of annual financial data for a specific company and correlates each year with the price of the stock. There are over 50 annual financial metrics used to correlate to the price during that year. This data is then stored as pickles and this process repeats for over 1000 companies, once there is enough data that can correlate financial metrics to price, all this data gets cleaned and converted to a table (Data Frames).

This data is furthermore checked again to make sure that the information is relatable to price and any uncorrelated data would be removed. Afterward, all this data is fed into a feed-forward neural network where the network gets trained and produce a model sufficient to predict results based on specific financial metrics input. Currently, the CSV being used is a list of financial metrics for a company, in this case, Apple. This data is then fed into a feed-forward neural network where the trained model is being used and the weights have already been adjusted to produce the correct result.

In this project case shown currently, the training would repeat for Apple, where it would quickly train for 100 Epochs, and the result would be the prediction price and the calculated Loss and MAE. A decreasing loss means that the network was able to effectively handle training and the error difference is getting lower, which means that the training within the network is working as intended. Please allow some time when running the algorithm as it trains from scratch when it runs. Please Note that the current result will not be very accurate as this is new training done and the model is not optimized to produce accurate result at 100 epochs only, however this was done using a powerful model which produced much better results.