Overview

This project predicts Google's stock prices using LSTM models built with Keras and PyTorch. The process includes data collection, preprocessing, feature engineering, and model evaluation.

Tech Stack

Challenges

• Handling missing values and scaling data for effective model performance.
• Feature engineering, selecting relevant features to improve model accuracy.
• Model selection and hyperparameter tuning for optimal performance.
• Preventing overfitting to ensure generalization to unseen data.

Solution

Missing values were handled and data was scaled using MinMaxScaler. SMAs and daily returns were added as features. LSTM models were built using Keras and PyTorch, and overfitting was tackled using dropout and early stopping.

Outcome

The models achieved reasonable accuracy in predicting future stock prices, with visualizations such as line charts and candlestick charts showing the effectiveness of the prediction. The project provided valuable insights into stock price forecasting and the potential of LSTM models.