app.py

import streamlit as st
import requests
import pandas as pd
from datetime import datetime, timedelta
import nltk
import numpy as np
from sklearn.linear_model import LinearRegression
import plotly.graph_objects as go
from plotly.subplots import make_subplots
import yfinance as yf
import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline

# --------------------------
# CONFIG
# --------------------------
st.set_page_config(page_title="📰 News Sentiment Analysis for Young Investor", layout="wide")
API_KEY = "88bc396d4eab4be494a4b86ec842db47"

# --------------------------
# โหลด FinBERT model
# --------------------------
@st.cache_resource
def load_finbert():
    tokenizer = AutoTokenizer.from_pretrained("project-aps/finbert-finetune")
    model = AutoModelForSequenceClassification.from_pretrained("project-aps/finbert-finetune")
    return tokenizer, model

tokenizer, model = load_finbert()

# --------------------------
# โหลด Zero-shot classifier สำหรับธีมข่าว
# --------------------------
@st.cache_resource
def load_theme_classifier():
    return pipeline("zero-shot-classification", model="facebook/bart-large-mnli")

theme_classifier = load_theme_classifier()
candidate_labels = ["Stock Movement", "Earnings", "M&A", "Regulation", "Product Launch", "Market Analysis"]

# --------------------------
# โหลด Pegasus สำหรับสรุปข่าว
# --------------------------
@st.cache_resource
def load_summarizer():
    return pipeline("summarization", model="Nerdward/financial-summarization-pegasus-finetuned-pytorch-model")

summarizer = load_summarizer()

# --------------------------
# UTILITIES
# --------------------------
def analyze_text(text):
    """วิเคราะห์อารมณ์ของข่าวด้วย FinBERT"""
    if not text or not text.strip():
        return 0

    inputs = tokenizer(
        text,
        return_tensors="pt",
        padding=True,
        truncation=True,
        max_length=512
    )

    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
        probs = torch.softmax(logits, dim=1).numpy()[0]

    # FinBERT = [negative, neutral, positive]
    score = (-1 * probs[0]) + (0 * probs[1]) + (1 * probs[2])
    return float(score)

def summarize_texts(news_texts):
    """สรุปข่าวแต่ละข่าว 1 พารากราฟ"""
    summaries = []
    for text in news_texts:
        if not text.strip():
            summaries.append("")
            continue
        try:
            summary = summarizer(text, max_length=100, min_length=30, do_sample=False)[0]["summary_text"]
            summaries.append(summary)
        except:
            summaries.append(text)  # fallback ถ้าโมเดลล้ม
    return summaries

def summarize_themes(news_texts):
    """สรุปธีมข่าวด้วย Zero-shot classification"""
    themes = []
    for text in news_texts:
        if not text.strip():
            themes.append("Unknown")
            continue
        result = theme_classifier(text, candidate_labels)
        themes.append(result["labels"][0])
    return themes

# --------------------------
# แปลงชื่อ/ตัวย่อ → (Company Name, Symbol)
# --------------------------
def resolve_company_symbol(keyword: str):
    keyword = keyword.strip()
    ticker = None
    name = None
    try:
        data = yf.Ticker(keyword)
        info = data.info
        if "symbol" in info and info["symbol"]:
            ticker = info["symbol"]
            name = info.get("longName", info.get("shortName", keyword))
        else:
            url = f"https://query2.finance.yahoo.com/v1/finance/search?q={keyword}"
            res = requests.get(url).json()
            if "quotes" in res and len(res["quotes"]) > 0:
                q = res["quotes"][0]
                ticker = q.get("symbol")
                name = q.get("longname", q.get("shortname", keyword))
    except:
        pass
    if not ticker:
        ticker = keyword.upper()
    if not name:
        name = keyword.capitalize()
    return name, ticker

# --------------------------
# ดึงข่าว 7 วัน
# --------------------------
@st.cache_data(ttl=3600)
def fetch_financial_news(keyword):
    company, symbol = resolve_company_symbol(keyword)
    to_date = datetime.now().strftime('%Y-%m-%d')
    from_date = (datetime.now() - timedelta(days=7)).strftime('%Y-%m-%d')
    query_keyword = f"({company} OR {symbol}) finance stock"

    all_articles = []
    page = 1
    while True:
        url = (
            f"https://newsapi.org/v2/everything?"
            f"q={query_keyword}&"
            f"from={from_date}&to={to_date}&"
            f"language=en&sortBy=publishedAt&"
            f"pageSize=100&page={page}&apiKey={API_KEY}"
        )
        r = requests.get(url)
        data = r.json()
        if data.get("status") != "ok":
            st.error(f"API Error: {data}")
            break

        articles = data.get("articles", [])
        if not articles:
            break

        for a in articles:
            if a["description"]:
                all_articles.append({
                    "date": pd.to_datetime(a["publishedAt"]),
                    "text": f"{a['title']} {a['description']}",
                    "source": a["source"]["name"],
                    "url": a["url"]
                })

        if len(articles) < 100:
            break
        page += 1

    return pd.DataFrame(all_articles)

# --------------------------
# ดึงราคาหุ้น
# --------------------------
@st.cache_data(ttl=3600)
def fetch_stock_price(symbol, start_date, end_date):
    try:
        start_str = (start_date - timedelta(days=2)).strftime('%Y-%m-%d')
        end_str = (end_date + timedelta(days=1)).strftime('%Y-%m-%d')
        df = yf.download(symbol, start=start_str, end=end_str, interval="1d")
        if df.empty:
            st.warning("ไม่พบข้อมูลราคาหุ้น")
            return pd.DataFrame()
        df = df.reset_index()
        df_subset = df[['Date', 'Close']]
        df_subset.columns = ['date', 'price']
        df_subset["date"] = pd.to_datetime(df_subset["date"].dt.date)
        return df_subset
    except Exception as e:
        st.warning(f"ดึงราคาหุ้นล้มเหลว: {e}")
        return pd.DataFrame()

# --------------------------
# MAIN APP
# --------------------------
def main():
    st.title("📰 News Sentiment Analysis for Young Investor")
    st.markdown("วิเคราะห์แนวโน้มอารมณ์ของข่าวย้อนหลัง 7 วัน พร้อมราคาหุ้น")

    # Sidebar
    with st.sidebar:
        keyword = st.text_input("ค้นหา Stock Symbol (เช่น AAPL, TSLA):", "")
        analyze_btn = st.button("วิเคราะห์เลย")

    if not analyze_btn:
        st.info("กรอกคำค้นแล้วกด 'วิเคราะห์เลย'")
        return

    # ดึงข่าว
    st.info(f"กำลังดึงข่าวย้อนหลัง 7 วันสำหรับ '{keyword}'...")
    news_df = fetch_financial_news(keyword)
    if news_df.empty:
        st.warning("ไม่พบบทความข่าว")
        return

    # วิเคราะห์ Sentiment
    st.info("กำลังวิเคราะห์อารมณ์ของข่าว...")
    news_df["sentiment"] = news_df["text"].apply(analyze_text)
    news_df["date"] = pd.to_datetime(news_df["date"])

    # สรุปข่าวเป็น 1 พารากราฟ
    st.info("กำลังสรุปเนื้อหาข่าว...")
    news_df["text"] = summarize_texts(news_df["text"].tolist())

    # Metrics
    avg_sentiment = news_df["sentiment"].mean()
    pos_pct = (news_df["sentiment"] > 0.1).mean() * 100
    neg_pct = (news_df["sentiment"] < -0.1).mean() * 100

    col1, col2, col3 = st.columns(3)
    col1.metric("ค่าเฉลี่ยอารมณ์ข่าว", f"{avg_sentiment:.2f}")
    col2.metric("ข่าวเชิงบวก", f"{pos_pct:.1f}%")
    col3.metric("ข่าวเชิงลบ", f"{neg_pct:.1f}%")

    # ธีมข่าว
    st.subheader("📰 ธีมข่าว (Top Theme per Article)")
    news_df["theme"] = summarize_themes(news_df["text"].tolist())
    theme_counts = news_df["theme"].value_counts()
    st.bar_chart(theme_counts)

    # ... ส่วนกราฟ Sentiment & Price เหมือนเดิม (ข้ามตรงนี้เพื่อไม่ให้ยาวเกินไป) ...

    # แสดงรายการข่าว
    st.subheader("📰 รายการข่าวทั้งหมด")
    st.dataframe(news_df[["date", "source", "text", "sentiment", "theme", "url"]], use_container_width=True)

# ---------------------------------------------------------
# RUN APP
# ---------------------------------------------------------
if __name__ == "__main__":
    nltk.download("stopwords", quiet=True)
    main()