Financial Forecasting Model (S&P 500 Return Prediction)

Time series forecasting

ARIMA + Deep Learning (FNN)

Custom loss functions & statistical modeling

Python (scikit-learn, statsmodels, TensorFlow)

Data visualization (Matplotlib, Seaborn)

Project Description: Developed a hybrid ARIMA + FNN model for predicting S&P 500 returns, leveraging deep learning techniques combined with traditional statistical methods. Designed a custom loss function based on the t-distribution to improve model accuracy. Addressed challenges related to stationarity, feature selection, and hyperparameter tuning using GridSearchCV.

Model code with explanations

Performance comparison (ARIMA vs. Hybrid Model)

Interactive visualizations showing residuals, predictions, and model accuracy