The Linear Regression model was trained using historical Google stock data, utilizing key features like past stock prices to predict future stock prices. The model achieved an impressive R-squared score of 0.9997, indicating an almost perfect fit for the training data. The Mean Squared Error (MSE) was around 0.3042, reflecting an exceptionally low error between the actual and predicted prices. This shows that the Linear Regression model effectively captures overall price trends but may not fully capture short-term volatility. The plot compares actual Google stock prices against predicted prices to visualize its performance.

The Random Forest model produced outstanding results with an R-squared score of 0.9991 and a Mean Squared Error (MSE) of 0.8181. This model performed exceptionally well in capturing the overall trend of Google stock prices, utilizing the ensemble of decision trees to reduce variance and improve prediction accuracy. It is more robust in handling noise in stock data and better at predicting stock prices during volatile periods compared to the Linear Regression model.

Gradient Boosting outperformed other models with an R-squared score of 0.9993 and a Mean Squared Error (MSE) of 0.6612, making it the most accurate model for predicting Google stock prices. By building models sequentially and minimizing prediction error, Gradient Boosting is particularly effective in capturing both long-term trends and short-term fluctuations in stock prices. The Gradient Boosting plot shows the close alignment between actual and predicted prices.

The LSTM Neural Network, designed to capture time-dependent patterns in stock prices, achieved an R-squared score of 0.9604 and a Mean Squared Error (MSE) of 35.3952. While it captured long-term trends well, the model struggled to maintain accuracy during periods of high volatility. Despite the lower performance compared to the other models, the LSTM still offers valuable insights into time-series forecasting and the potential of deep learning for stock prediction.