fierce-dashboard

byKanchan Negi

Act as a Senior Machine Learning Engineer, Data Scientist, and Expert UI/UX Designer. Your task is to build a complete end-to-end project titled: "Intelligent Stock Market Forecasting Using LSTM With Interactive Dashboard And Chatbot Support" The implementation MUST strictly follow the system architecture, methodology, and objectives described below. Do not add extra features beyond this specification. ━━━━━━━━━━━━━━━━━━━━━━━ 🎯 PROJECT OVERVIEW ━━━━━━━━━━━━━━━━━━━━━━━ Build a Streamlit-based web application that predicts stock prices using an LSTM model trained on historical stock data. The system must include: 1. Data collection from Yahoo Finance 2. Data preprocessing and feature engineering 3. LSTM-based prediction model 4. Technical indicators visualization 5. Interactive dashboard (premium UI) 6. Rule-based chatbot for user queries The system should replicate a modern AI-powered financial dashboard and also act as an educational tool. ━━━━━━━━━━━━━━━━━━━━━━━ 🎨 PREMIUM UI/UX REQUIREMENTS (IMPORTANT) ━━━━━━━━━━━━━━━━━━━━━━━ Design the Streamlit dashboard to look like a modern trading platform (clean, dark, professional UI): * Use a DARK THEME (black/charcoal background) * Use custom CSS for styling (via st.markdown) * Layout should be: * Top header with title + subtitle * Sidebar for controls * Main dashboard with cards and sections Dashboard design must include: ✔ Metric Cards (at top): * Current Price * Predicted Price * % Change * RMSE / MAE ✔ Graph Section: * Large interactive price chart (Plotly) * Smooth lines, hover tooltips, zoom enabled ✔ Tabs or Sections: * Overview * Technical Indicators * Predictions * Chatbot ✔ Indicators Styling: * Colored lines (SMA 20 = blue, SMA 50 = orange) * RSI with threshold lines (30, 70) * MACD with histogram bars ✔ Layout Features: * Use columns for clean spacing * Rounded containers / card-like sections * Add spacing and padding for readability ✔ Chatbot UI: * Chat bubbles (user vs bot) * Scrollable chat history * Input box at bottom The UI should feel like: 👉 TradingView / Zerodha Kite / modern fintech dashboard ━━━━━━━━━━━━━━━━━━━━━━━ 📊 DATA COLLECTION ━━━━━━━━━━━━━━━━━━━━━━━ * Use yfinance to fetch historical stock data * Input: user-entered stock ticker (e.g., TCS.NS, RELIANCE.NS) * Allow user to select: * Start date * End date * Fetch OHLCV data ━━━━━━━━━━━━━━━━━━━━━━━ 🧹 DATA PREPROCESSING ━━━━━━━━━━━━━━━━━━━━━━━ * Handle missing values using forward fill * Normalize data using MinMaxScaler (0 to 1) * Train-test split: 80/20 ━━━━━━━━━━━━━━━━━━━━━━━ 📈 FEATURE ENGINEERING ━━━━━━━━━━━━━━━━━━━━━━━ Compute: 1. Moving Averages: * 20-day SMA * 50-day SMA 2. RSI (14-day) 3. MACD: * EMA(12) - EMA(26) * Signal line (9-day EMA) ━━━━━━━━━━━━━━━━━━━━━━━ 🧠 LSTM MODEL ━━━━━━━━━━━━━━━━━━━━━━━ * Sequence length: 60 * 2 LSTM layers (50 units each) * Dense output layer * Optimizer: Adam * Loss: MSE * Epochs: 50–100 * Batch size: 32 * Use early stopping ━━━━━━━━━━━━━━━━━━━━━━━ 🔮 PREDICTION ━━━━━━━━━━━━━━━━━━━━━━━ * Predict next 30 days * Use recursive forecasting * Inverse transform output ━━━━━━━━━━━━━━━━━━━━━━━ 📊 PERFORMANCE METRICS ━━━━━━━━━━━━━━━━━━━━━━━ Display: * RMSE * MAE ━━━━━━━━━━━━━━━━━━━━━━━ 🖥️ STREAMLIT DASHBOARD ━━━━━━━━━━━━━━━━━━━━━━━ Sidebar: * Stock ticker input * Start date * End date * Run Forecast button Main UI: 1. Overview: * Metric cards * Closing price chart 2. Technical Indicators: * SMA chart * RSI chart * MACD chart 3. Predictions: * Historical + Forecasted chart 4. Model Training Info (optional simple progress display) 5. Performance Metrics section ━━━━━━━━━━━━━━━━━━━━━━━ 💬 CHATBOT (NO API) ━━━━━━━━━━━━━━━━━━━━━━━ * Rule-based chatbot * Keyword matching * Answer: * LSTM * RSI * MACD * Moving average * Prediction meaning * Maintain conversation history ━━━━━━━━━━━━━━━━━━━━━━━ 📁 PROJECT STRUCTURE ━━━━━━━━━━━━━━━━━━━━━━━ * app.py * data_loader.py * preprocessing.py * indicators.py * model.py * predict.py * chatbot.py * styles.css (for UI) ━━━━━━━━━━━━━━━━━━━━━━━ ⚠️ CONSTRAINTS ━━━━━━━━━━━━━━━━━━━━━━━ DO NOT include: * Authentication * Database * Real-time streaming * Paid APIs * React or frontend frameworks Use ONLY: Python, Streamlit, yfinance, pandas, numpy, scikit-learn, TensorFlow/Keras, Plotly ━━━━━━━━━━━━━━━━━━━━━━━ 🎯 FINAL OUTPUT ━━━━━━━━━━━━━━━━━━━━━━━ * Fully working Streamlit app * Premium UI dashboard * Clean modular code * requirements.txt included * Step-by-step run instructions Ensure the project is visually impressive and suitable for final-year B.Tech evaluation.