PSE Stock Dashboard

After months of dedicated work and countless hours invested, this web-based Streamlit application is a PSE stock dashboard designed to predict the next day’s price movement. At its core, the app leverages a tree-based machine learning model trained on an extensive set of features, including opening, high, low, and closing prices, trading volume, sentiment scores (positive, negative, and neutral), PSE Edge disclosures, and several technical indicators such as SMA 20, EMA 12, EMA 26, MACD, signal line, MACD histogram, and RSI. 

Sentiment features are extracted from company disclosures published on PSE Edge using FinBERT via Huggingface API calls, ensuring that the model captures market sentiment effectively. The app offers a suite of user-friendly features, including a sidebar for time series configuration, dataset visualization, and options for selecting features to plot. 

The interface is organized into three main tabs: Prediction, Plots, and SHAP Analysis. The Prediction tab enables users to input values such as open, high, low, close, volume, sentiment, and disclosure data. It also utilizes the OpenAI API (with gpt-4o-mini as the model) to deliver AI-powered stock analyses, including buy and sell rationales and a comprehensive SWOT analysis. 

The Plots tab provides interactive visualizations, including time series charts, tabular dataset displays, price movement trends, annual returns, standard deviation, and risk-return profiles. The SHAP Analysis tab enhances model transparency by offering interpretability tools that help users understand the factors influencing the model’s predictions.