A No-Code Machine Learning System for Next Day Equity and Crypto Forecasting Experiments

As machine learning modeling becomes more robust and codified with well known and proven software packages such as SciKit Learn and Tensorflow, the fact remains that serious coding, data cleaning and prepping, and machine learning knowledge is required to use these tools effectively. Here we present a system usable by retail traders without any coding required which allows users to enter stock ticker symbols or crypto currency symbols into a small software application ("app") to kick off a machine learning and backtesting process. The purpose: to try to forecast the next day's price change of one asset and therefore ignoring or attempting to circumvent the stronger forms of the Efficient Market Hypothesis. The system is set up to automatically pre-process data to generate regression tables, generate base models, backtest these models over a user-specified historical window, then present only relevant backtest metrics for analysis. The app then allows basic manual tuning of models to improve backtests by providing a small click and tap user interface with some (but not all conceivable) model parameters exposed to the user. Finally, when the user is satisfied with backtests, with one menu click, the system can be set to forecast the "next" day's price travel of the target asset