In the context of the Genius Hour Project in a marketing class at Esquimalt High School, we decided to embark on an ambitious challenge: automating trading and stock market investments. Over three months, we worked, adjusted, and tested our bot. Frederike handled the coding, aiming first for the bot to analyze various technical data and respond whether or not we should buy the stock. Meanwhile, I focused on finding a browser that would give us access to virtual money and a free API that could be integrated with the AI.
Following this, the bot was improved and became capable of directly purchasing a stock with a positive outcome. It provided the growth percentage of a stock, ranging from 1% to over 800%—a bit unrealistic for a first version that analyzed only a limited amount of data. Then, we started investing. The bot analyzed over 500 U.S. stocks and suggested 50 with the highest growth percentages. A week later, the initial $100,000 counter showed a +1.5% increase. This pleasing result allowed us to sell half and wait a bit longer until we achieved a $2,000 profit.
For this version, we essentially focused on improving decision-making and the questions the bot needs to ask itself before investing. To achieve this, we needed to enhance our AI model layers with hidden layers.
This refers to the architecture of artificial neural networks, where hidden layers process data between the input layer and the output layer. Hidden layers enable a model to learn complex and non-linear features from the data.
Basically, we transitioned from a table with 34 million branches to a GRU Keras model, which handles more complex patterns, even though processing this amount of information takes more time to analyze.
This version actively manages a paper portfolio. Before the pre market hours the AI analysises all 3.254 stocks in the Wilshire 3000 index. Then stocks with a good prediction will be baught. The price for what is determined by the predicition and the available cash. On every market closing hour the AI analisies the porftfolios stocks and determines to sell them or not. Also a top price order is placed when buying containing as the stop price the predicted growth, so the stock imdeadly gets sold when the stock price rasises to the predicted value.