import streamlit as st import pandas as pd import yfinance as yf from textblob import TextBlob import joblib import matplotlib.pyplot as plt from datetime import datetime import plotly.express as px # Function to load stock data using yfinance/ this is going to refresh after 1 day @st.cache_data(ttl=86400) def load_stock_data(tickers, start_date, end_date): with st.spinner('Downloading stock data...'): data = yf.download(tickers, start=start_date, end=end_date, group_by='ticker', auto_adjust=True) all_data = [] for ticker in tickers: df = data[ticker].copy().reset_index() df['Stock Name'] = ticker all_data.append(df) merged_data = pd.concat(all_data, ignore_index=True) return merged_data tickers = ['TSLA', 'MSFT', 'PG', 'META', 'AMZN', 'GOOG', 'AMD', 'AAPL', 'NFLX', 'TSM', 'KO', 'F', 'COST', 'DIS', 'VZ', 'CRM', 'INTC', 'BA', 'BX', 'NOC', 'PYPL', 'ENPH', 'NIO', 'ZS', 'XPEV'] start_date = (datetime.today() - pd.DateOffset(years=1)).strftime('%Y-%m-%d') end_date = datetime.today().strftime('%Y-%m-%d') # Cache stock data for 1 day using st.cache_data @st.cache_data(ttl=86400) def load_and_cache_stock_data(): return load_stock_data(tickers, start_date, end_date) # Initialize stock_data once at app startup if "stock_data" not in st.session_state: st.session_state["stock_data"] = load_and_cache_stock_data() stock_data = st.session_state["stock_data"] # Perform sentiment analysis on tweets (assuming you still have your tweets data) tweets_data = pd.read_csv('data/stock_tweets.csv') # Convert the Date columns to datetime tweets_data['Date'] = pd.to_datetime(tweets_data['Date']).dt.date # Perform sentiment analysis on tweets def get_sentiment(tweet): analysis = TextBlob(tweet) return analysis.sentiment.polarity tweets_data['Sentiment'] = tweets_data['Tweet'].apply(get_sentiment) # Aggregate sentiment by date and stock daily_sentiment = tweets_data.groupby(['Date', 'Stock Name']).mean(numeric_only=True).reset_index() # Convert the Date column in daily_sentiment to datetime64[ns] daily_sentiment['Date'] = pd.to_datetime(daily_sentiment['Date']) # Merge stock data with sentiment data merged_data = pd.merge(stock_data, daily_sentiment, how='left', on=['Date', 'Stock Name']) # Fill missing sentiment values with 0 (neutral sentiment) merged_data['Sentiment'] = merged_data['Sentiment'].fillna(0) # Sort the data by date merged_data = merged_data.sort_values(by='Date') # Create lagged features merged_data['Prev_Close'] = merged_data.groupby('Stock Name')['Close'].shift(1) merged_data['Prev_Sentiment'] = merged_data.groupby('Stock Name')['Sentiment'].shift(1) # Create moving averages merged_data['MA7'] = merged_data.groupby('Stock Name')['Close'].transform(lambda x: x.rolling(window=7).mean()) merged_data['MA14'] = merged_data.groupby('Stock Name')['Close'].transform(lambda x: x.rolling(window=14).mean()) # Create daily price changes merged_data['Daily_Change'] = merged_data['Close'] - merged_data['Prev_Close'] # Create volatility merged_data['Volatility'] = merged_data.groupby('Stock Name')['Close'].transform(lambda x: x.rolling(window=7).std()) # Drop rows with missing values merged_data.dropna(inplace=True) # Load the best model model_filename = 'model/best_model.pkl' model = joblib.load(model_filename) # Streamlit application layout st.title("Stock Price Prediction Using Sentiment Analysis") # User input for stock data st.header("Input Stock Data") stock_names = merged_data['Stock Name'].unique() selected_stock = st.selectbox("Select Stock Name", stock_names) days_to_predict = st.number_input("Number of Days to Predict", min_value=1, max_value=30, value=10) # Get the latest data for the selected stock latest_data = merged_data[merged_data['Stock Name'] == selected_stock].iloc[-1] prev_close = latest_data['Close'] prev_sentiment = latest_data['Sentiment'] ma7 = latest_data['MA7'] ma14 = latest_data['MA14'] daily_change = latest_data['Daily_Change'] volatility = latest_data['Volatility'] # Display the latest stock data in a table latest_data_df = pd.DataFrame({ 'Metric': ['Previous Close Price', 'Previous Sentiment', '7-day Moving Average', '14-day Moving Average', 'Daily Change', 'Volatility'], 'Value': [prev_close, prev_sentiment, ma7, ma14, daily_change, volatility] }) st.write("Latest Stock Data:") st.write(latest_data_df) st.write("Use the inputs above to predict the next days close prices of the stock.") if st.button("Predict"): predictions = [] latest_date = datetime.now() for i in range(days_to_predict): X_future = pd.DataFrame({ 'Prev_Close': [prev_close], 'Prev_Sentiment': [prev_sentiment], 'MA7': [ma7], 'MA14': [ma14], 'Daily_Change': [daily_change], 'Volatility': [volatility] }) next_day_prediction = model.predict(X_future)[0] predictions.append(next_day_prediction) # Update features for next prediction prev_close = next_day_prediction ma7 = (ma7 * 6 + next_day_prediction) / 7 # Simplified rolling calculation ma14 = (ma14 * 13 + next_day_prediction) / 14 # Simplified rolling calculation daily_change = next_day_prediction - prev_close # Prepare prediction data for display prediction_dates = pd.date_range(start=latest_date + pd.Timedelta(days=1), periods=days_to_predict) prediction_df = pd.DataFrame({ 'Date': prediction_dates, 'Predicted Close Price': predictions }) st.subheader("Predicted Prices") st.dataframe(prediction_df) # Plot predictions using Plotly fig = px.line(prediction_df, x='Date', y='Predicted Close Price', markers=True, title=f"{selected_stock} Predicted Close Prices") st.plotly_chart(fig, use_container_width=True) # ---------------------------------------- # Enhanced Visualizations st.header("Enhanced Stock Analysis") stock_history = merged_data[merged_data['Stock Name'] == selected_stock] # Date filter slider min_date = pd.to_datetime(merged_data['Date'].min()).date() max_date = pd.to_datetime(merged_data['Date'].max()).date() date_range = st.slider( "Select Date Range for Visualizations", min_value=min_date, max_value=max_date, value=(min_date, max_date), format="YYYY-MM-DD" ) filtered_data = stock_history[(stock_history['Date'] >= pd.to_datetime(date_range[0])) & (stock_history['Date'] <= pd.to_datetime(date_range[1]))] with st.expander("Price vs Sentiment Trend"): fig1 = px.line(filtered_data, x='Date', y=['Close', 'Sentiment'], labels={'value': 'Price / Sentiment', 'variable': 'Metric'}, title=f"{selected_stock} - Close Price & Sentiment") st.plotly_chart(fig1, use_container_width=True) with st.expander("Volatility Over Time"): fig2 = px.line(filtered_data, x='Date', y='Volatility', title=f"{selected_stock} - 7-Day Rolling Volatility") st.plotly_chart(fig2, use_container_width=True) with st.expander("Moving Averages (MA7 vs MA14)"): fig3 = px.line(filtered_data, x='Date', y=['MA7', 'MA14'], labels={'value': 'Price', 'variable': 'Moving Average'}, title=f"{selected_stock} - Moving Averages") st.plotly_chart(fig3, use_container_width=True) with st.expander("Daily Price Change"): fig4 = px.line(filtered_data, x='Date', y='Daily_Change', title=f"{selected_stock} - Daily Price Change") st.plotly_chart(fig4, use_container_width=True) with st.expander("Sentiment Distribution"): fig5 = px.histogram(filtered_data, x='Sentiment', nbins=30, title=f"{selected_stock} - Sentiment Score Distribution") st.plotly_chart(fig5, use_container_width=True)