app.py

import streamlit as st
from urllib.request import urlopen, Request
from bs4 import BeautifulSoup
import pandas as pd
import plotly.express as px
from dateutil import parser
import datetime
import requests
from transformers import BertTokenizer, BertForSequenceClassification, pipeline

st.set_page_config(page_title="Stock News Sentiment Analyzer", layout="wide")

# Initialize FinBERT (yiyanghkust/finbert-tone) pipeline only once and cache
@st.cache_resource(show_spinner=False)
def load_model():
    model = BertForSequenceClassification.from_pretrained('yiyanghkust/finbert-tone', num_labels=3)
    tokenizer = BertTokenizer.from_pretrained('yiyanghkust/finbert-tone')
    return pipeline("sentiment-analysis", model=model, tokenizer=tokenizer)

finbert = load_model()

def verify_link(url, timeout=10, retries=3):
    for _ in range(retries):
        try:
            response = requests.head(url, timeout=timeout, allow_redirects=True)
            if 200 <= response.status_code < 300:
                return True
        except requests.RequestException:
            continue
    return False

def get_news(ticker):
    try:
        finviz_url = 'https://finviz.com/quote.ashx?t='
        url = finviz_url + ticker
        req = Request(url=url, headers={'User-Agent': 'Mozilla/5.0'}) 
        response = urlopen(req)    
        html = BeautifulSoup(response, 'html.parser')
        news_table = html.find(id='news-table')
        return news_table
    except Exception as e:
        st.write("Error fetching news:", str(e))
        return None

def parse_news(news_table):
    parsed_news = []
    
    for x in news_table.findAll('tr'):
        try:
            text = x.a.get_text()
            link = x.a['href']
            date_scrape = x.td.text.strip().split()
            
            if len(date_scrape) == 1:
                date = datetime.datetime.today().strftime('%Y-%m-%d')
                time = date_scrape[0]
            else:
                date = date_scrape[0]
                time = date_scrape[1]

            datetime_str = f"{date} {time}"
            datetime_parsed = parser.parse(datetime_str)

            is_valid = verify_link(link)
            
            parsed_news.append([datetime_parsed, text, link, is_valid])
            
        except Exception as e:
            print("Error parsing news:", e)
            continue
    
    columns = ['datetime', 'headline', 'link', 'is_valid']
    parsed_news_df = pd.DataFrame(parsed_news, columns=columns)
    
    return parsed_news_df

def score_news(parsed_news_df):
    # Send headlines in smaller batches to speed up processing
    batch_size = 10
    parsed_news_df = parsed_news_df.reset_index(drop=True)
    sentiment_scores = []

    for i in range(0, len(parsed_news_df), batch_size):
        batch_headlines = parsed_news_df['headline'][i:i+batch_size].tolist()
        predictions = finbert(batch_headlines)
        
        for pred in predictions:
            label = pred['label']
            score = pred['score']
            
            # Map 'LABEL_0' (neutral), 'LABEL_1' (positive), and 'LABEL_2' (negative)
            if label == 'LABEL_1':
                sentiment_score = score
            elif label == 'LABEL_2':
                sentiment_score = -score
            else:  # neutral
                sentiment_score = 0
                
            sentiment_scores.append({
                'sentiment_score': sentiment_score,
                'label': label,
                'confidence': score
            })

    scores_df = pd.DataFrame(sentiment_scores)
    parsed_and_scored_news = pd.concat([parsed_news_df, scores_df], axis=1)
    parsed_and_scored_news = parsed_and_scored_news.set_index('datetime')
    
    return parsed_and_scored_news

def plot_hourly_sentiment(parsed_and_scored_news, ticker):
    mean_scores = parsed_and_scored_news['sentiment_score'].resample('h').mean()
    
    fig = px.bar(mean_scores, x=mean_scores.index, y='sentiment_score', 
                 title=f'{ticker} Hourly Sentiment Scores',
                 color='sentiment_score',
                 color_continuous_scale=['red', 'yellow', 'green'],
                 range_color=[-1, 1])
    
    fig.update_layout(coloraxis_colorbar=dict(
        title="Sentiment",
        tickvals=[-1, 0, 1],
        ticktext=["Negative", "Neutral", "Positive"],
    ))
    
    return fig

def plot_daily_sentiment(parsed_and_scored_news, ticker):
    mean_scores = parsed_and_scored_news['sentiment_score'].resample('D').mean()
    
    fig = px.bar(mean_scores, x=mean_scores.index, y='sentiment_score', 
                 title=f'{ticker} Daily Sentiment Scores',
                 color='sentiment_score',
                 color_continuous_scale=['red', 'yellow', 'green'],
                 range_color=[-1, 1])
    
    fig.update_layout(coloraxis_colorbar=dict(
        title="Sentiment",
        tickvals=[-1, 0, 1],
        ticktext=["Negative", "Neutral", "Positive"],
    ))
    
    return fig

def get_recommendation(sentiment_scores):
    avg_sentiment = sentiment_scores['sentiment_score'].mean()
    
    if avg_sentiment >= 0.3:
        return f"Positive sentiment (Score: {avg_sentiment:.2f}). The recent news suggests a favorable outlook for this stock. Consider buying or holding if you already own it."
    elif avg_sentiment <= -0.3:
        return f"Negative sentiment (Score: {avg_sentiment:.2f}). The recent news suggests caution. Consider selling or avoiding this stock for now."
    else:
        return f"Neutral sentiment (Score: {avg_sentiment:.2f}). The recent news doesn't show a strong bias. Consider holding if you own the stock, or watch for more definitive trends before making a decision."

st.header("Stock News Sentiment Analyzer (FinBERT)")

ticker = st.text_input('Enter Stock Ticker', '').upper()

try:
    st.subheader(f"Sentiment Analysis and Recommendation for {ticker} Stock")
    news_table = get_news(ticker)
    if news_table:
        parsed_news_df = parse_news(news_table)
        parsed_and_scored_news = score_news(parsed_news_df)

        # Generate and display recommendation
        recommendation = get_recommendation(parsed_and_scored_news)
        st.write(recommendation)
        
        # Display a disclaimer
        st.warning("Disclaimer: This recommendation is based solely on recent news sentiment and should not be considered as financial advice. Always do your own research and consult with a qualified financial advisor before making investment decisions.")
        
        fig_hourly = plot_hourly_sentiment(parsed_and_scored_news, ticker)
        fig_daily = plot_daily_sentiment(parsed_and_scored_news, ticker) 
        
        st.plotly_chart(fig_hourly)
        st.plotly_chart(fig_daily)

        description = f"""
            The above charts average the sentiment scores of {ticker} stock hourly and daily.
            The table below shows recent headlines with their sentiment scores and classifications.
            The news headlines are obtained from the FinViz website.
            Sentiments are analyzed using the ProsusAI/finbert model, which is specifically trained for financial text.
            Links have been verified for validity.
            """
            
        st.write(description)     
        
        parsed_and_scored_news['link'] = parsed_and_scored_news.apply(
            lambda row: f'<a href="{row["link"]}" target="_blank">{"Valid✅" if row["is_valid"] else "Invalid❌"} Link</a>', 
            axis=1
        )
        
        display_df = parsed_and_scored_news.drop(columns=['is_valid'])
        st.write(display_df.to_html(escape=False), unsafe_allow_html=True)

    else:
        st.write("No news available or invalid ticker symbol.")
    
except Exception as e:
    print(str(e))
    st.write("Enter a correct stock ticker, e.g. 'AAPL' above and hit Enter.")

hide_streamlit_style = """
<style>
#MainMenu {visibility: hidden;}
footer {visibility: hidden;}
</style>
"""
st.markdown(hide_streamlit_style, unsafe_allow_html=True)