In creation of this project, we decided to pursue a mission to provide more accessible and easy to use data, without having to pay for it. This incentivizes more and more people to invest in the stock market, allowing companies to gain the necessary money on a need basis, and everyone to enjoy long term capital gains with their holds in the market. This proves to be a win-win for everyone, as current investors, new investors, and companies will all benefit and make more money with more and new investors in the markets.
What it does
Our application prompts the user for a ticker symbol they want to investigate, with an automatic completion system while searching for a ticker. This is then processed to our server, then to our stock analysis code, and results are displayed as a response in the UI/UX. This machine learning service has over a 94.7% accuracy on it's trained models. It takes data over 10 years and splits into training and testing, and effectively gives the user a nicely time stamped prediction for future prices based on adjusted close price.
How we built it
We used Flutter to build the UI/UX to allow iOS and AndroidOS to both have access to the application. We used Flask and the requests library to process requests and responses to and from the server. Finally, we used Python, and within python, we used the TensorFlow library to complete machine learning and data analysis on the stock's adjusted close price. After using training data and testing data from the past ten years, we are able to effectively predict and display a graph for the current stock.
Challenges we ran into
Our biggest challenge was attempting to connect the server with the algorithm, as we got many errors trying to run the algorithm as a function. We still face some of those issues, but we have tried to suppress the errors as best as possible.
Accomplishments that we're proud of
We are very proud of our UI execution and server's HTTP networking, as well as our stock prediction algorithm. All were highly complex pieces of code with tons of errors, but some patience and Stack Overflow searches helped us reach the end product we reached currently.
What we learned
We all learned to troubleshoot errors, and how to create perfect architecture for HTTP requests and servers, to compute the most optimal server for the given project. The UI/UX taught us a lot about design and implementation, and the machine learning taught us the nature of errors and how to adjust code for other pieces of code.
What's next for Stock Price Prediction Mobile Application
Next, we hope to add an interactive graph, so you can see all future values, and then we hope to include more advanced algorithms and options to trade right in our application. This would take Stock Portfolio Allocation to the next level in terms of accessible algorithms for general stock traders and investors.