FIG1. Stock Prediction Model's Training vs Actual Data
S&P500
INTERACTIVE GRAPH
BLUE: Model Prediction
BLACK: Actual Price
Evaluation: The model has been trained on all the S&P500 data from 2000-2022. The model has some good guesses when there will be an uptrend or a down trend, but the actual values tend to overshoot or undershoot.
FIG2. Same Model's Prediction 6 months into the future
INTERACTIVE GRAPH
BLUE: Model Prediction
BLACK: Actual Price
Evaluation: It seems that the model doesn't know how to make sense of the most recent bull market (Oct 2023 - March 2024). Ideas for improvements: 1) Get more relevant data for uptrends and downtrends. 2) Add more weight to those data.
This is exciting because the model is already making pseudo intellectual guesses based on just 4 variables! Imagine if more quality data was fed to this model.
Technologies Used: Plotly, Plotly Express, Pandas, XGBOOST