a lot of changes

App/app.py

@@ -12,19 +12,22 @@ import os
 import plotly.express as px
 import sys
 
+print('Starting App...')
+
 PROJECT_ROOT = r"C:\Users\anime\OneDrive\Desktop\Sentiment-Analysis-App"
+# This block adds your project folder to Python's search path so it can find 'finance'.
 if PROJECT_ROOT not in sys.path:
     sys.path.insert(0, PROJECT_ROOT)
 
-# This brings in the logic from your new 'finance' directory
+# --- Import New Finance Modules ---
+# This will now work correctly because of the path correction above.
 from finance.processor import (
-    ingest_transcripts,
-    preprocess_and_split,
-    get_sentiment_vectors,
-    aggregate_vectors_to_features,
-    get_stock_returns
+    get_huggingface_sentiment,
+    get_twitter_credentials,
+    get_twitter_data,
 )
-from finance.analysis import run_prediction_model
+from finance.modeling_analysis import run_prediction_model
+
 
 # Configuration and Model Loading
 
@@ -38,35 +41,28 @@ st.set_page_config(
 
 @st.cache_resource
 def load_models():
-    """
-    Loads both the general emotion model and the fine-tuned financial model.
-    This function will be run only once.
-    """
+    """Loads both models using the hardcoded project root path."""
     models = {}
     try:
-        # --- Determine Project Root Correctly ---
-        # This handles running the script from 'App/' or the project's root directory
-        app_dir = os.path.dirname(os.path.abspath(__file__))
-        if os.path.basename(app_dir) == 'App':
-            project_root = os.path.dirname(app_dir)
-        else:
-            project_root = app_dir
-
-        # --- Load General Emotion Model (for the first tab) ---
-        general_model_path = os.path.join(project_root, "Models", "sentiment_model_distilbert")
+        # --- Load General Emotion Model ---
+        # This path is for the "Emotion Analyzer" tab.
+        # It should point to your trained DistilBERT model.
+        general_model_path = os.path.join(PROJECT_ROOT, "Models", "sentiment_model_distilbert")
         st.info(f"Loading general model from: {general_model_path}")
         models['general_tokenizer'] = AutoTokenizer.from_pretrained(general_model_path)
         models['general_model'] = AutoModelForSequenceClassification.from_pretrained(general_model_path)
         
-        # --- Load Fine-Tuned Financial Model (for the second tab) ---
-        finbert_path = os.path.join(project_root, "finance", "finbert_emotion_model")
+        # --- Load Fine-Tuned Financial Model ---
+        # This path is for the "Financial Analysis" tab.
+        # It should point to your final, high-accuracy model (e.g., finbert_large_emotion_model).
+        finbert_path = os.path.join(PROJECT_ROOT, "finance", "finbert_large_emotion_model") 
         st.info(f"Loading financial model from: {finbert_path}")
         if os.path.exists(finbert_path):
             models['finbert_tokenizer'] = AutoTokenizer.from_pretrained(finbert_path)
             models['finbert_model'] = AutoModelForSequenceClassification.from_pretrained(finbert_path)
         else:
-            # Display a warning if the financial model hasn't been trained yet
-            st.sidebar.warning("Fine-tuned FinBERT model not found. Please run `python finance/train_finbert.py` from your project root.")
+            # This warning will appear if the folder is not found
+            st.sidebar.warning("Fine-tuned FinBERT model not found. Please run the training script.")
             models['finbert_tokenizer'] = None
             models['finbert_model'] = None
 
@@ -74,7 +70,6 @@ def load_models():
         st.error(f"Error loading models: {e}")
         return None
     return models
-
 # Load all models at startup
 models = load_models()
 
@@ -156,7 +151,7 @@ def run_full_finance_pipeline():
     finbert_model = models['finbert_model']
     finbert_tokenizer = models['finbert_tokenizer']
     
-    df_transcripts = ingest_transcripts()
+    df_transcripts = ingest_transcripts(PROJECT_ROOT)
     all_features = []
 
     for _, row in df_transcripts.iterrows():
@@ -238,4 +233,6 @@ def render_about_page():
     """)
 
 if __name__ == '__main__':
-    main()
+    main()
+
+print("App loaded successfully.")

finance/advanced_earnings.csv

@@ -1,4 +0,0 @@
-ticker,company_name,quarter,year,earnings_date,prepared_remarks,analyst_qa
-AAPL,Apple Inc.,4,2023,2023-10-26,"Good afternoon everyone. We are thrilled to report a September quarter revenue record. Our services division achieved an all-time revenue peak, and iPhone sales were the strongest we've ever seen for this quarter. Our active installed base continues to grow to new heights, which is a testament to our customer loyalty. We are confident in our product pipeline and see strong momentum heading into the holiday season. Our margins remain healthy despite some supply chain headwinds. Overall, we executed extremely well.","Analyst: Can you comment on the regulatory scrutiny in Europe? Tim Cook: We are actively engaged with regulators and are confident we comply with all laws. It's a complex issue, but we are focused on innovation. Analyst: What is the outlook for your China market given the recent slowdown? Tim Cook: We saw some softness, which we attribute to macroeconomic factors. However, we remain very optimistic about the long-term opportunity in China. We believe our products offer tremendous value."
-MSFT,Microsoft Corp.,1,2024,2023-10-24,"Thank you for joining. We are off to a tremendous start to the fiscal year. The Microsoft Cloud surpassed $31.8 billion in quarterly revenue, a fantastic achievement. Our AI initiatives, especially Copilot, are driving significant productivity gains for our customers and creating new revenue streams. We are rapidly infusing AI across every layer of the tech stack to lead this new era. Bookings were strong, and we feel we are well-positioned to continue our growth trajectory. Our commercial business is performing exceptionally well.","Analyst: How should we think about the margin impact from your AI investments? Amy Hood: There is an upfront investment cost, but we expect the efficiencies and new revenue will lead to margin expansion over time. It's a strategic investment in our future. Analyst: Is the PC market showing signs of recovery? Satya Nadella: The market is stabilizing. We see challenges, but also pockets of strength. We are focused on innovating in the PC space with AI-powered experiences."
-GOOGL,Alphabet Inc.,3,2023,2023-10-24,"Good afternoon. I am very pleased with our financial results this quarter. Revenues were strong, up 11% year over year, driven by excellent performance in Search and YouTube, along with continued momentum in our Cloud division. Our investments in AI are bearing fruit, and our new Gemini models are performing at the state-of-the-art. We are laser-focused on delivering value for our users and advertisers. The future is bright, and we are innovating at a rapid pace.","Analyst: Cloud growth seems to have decelerated slightly. Can you provide color? Ruth Porat: We are still seeing very strong customer adoption. The growth rate can fluctuate quarter to quarter based on deal timing, but we are confident in the underlying business momentum. There is some concern, but we are managing it. Analyst: What is the timeline for monetizing your generative AI products in Search? Sundar Pichai: We are taking a measured approach. We are testing various models and user experiences. It is a long-term journey, and we are being very deliberate to ensure we get it right for our users and partners."

finance/processor.py

@@ -7,74 +7,100 @@ import yfinance as yf
 from datetime import timedelta
 import numpy as np
 import torch
+import os
+from sklearn.feature_extraction.text import TfidfVectorizer
 
+def get_sentiment_vectors(texts: list[str]) -> list[list[float]]:
+    """Return TF-IDF vectors for a list of texts."""
+    vectorizer = TfidfVectorizer(stop_words="english")
+    return vectorizer.fit_transform(texts).toarray()
 
-def ingest_transcripts(file_path='finance/sample_transcripts_advanced.csv'):
-    """Ingests earnings call transcripts from the advanced CSV file."""
-    return pd.read_csv(file_path)
-
+def ingest_transcripts(project_root: str):
+    """
+    Ingests the NEW, simplified sample earnings call transcripts.
+    """
+    file_path = os.path.join(project_root, "finance", "sample_transcripts.csv")
+    try:
+        return pd.read_csv(file_path)
+    except FileNotFoundError:
+        raise FileNotFoundError(f"Could not find transcript data at '{file_path}'.")
 
-def preprocess_and_split(transcript_text):
-    """Splits raw text into clean, individual sentences."""
-    if not isinstance(transcript_text, str):
+def preprocess_and_split(text):
+    """Cleans and splits text into sentences."""
+    if not isinstance(text, str):
         return []
-    sentences = re.split(r'(?<=[.!?]) +', transcript_text)
-    cleaned_sentences = [s.strip() for s in sentences if s.strip()]
-    return cleaned_sentences
+    text = re.sub(r'\s+', ' ', text).strip()
+    sentences = re.split(r'(?<=[.!?])\s+', text)
+    return [s.strip() for s in sentences if s.strip()]
 
-
-def get_sentiment_vectors(sentences, model, tokenizer):
-    """Runs FinBERT on sentences to get full emotion probability vectors."""
+@torch.no_grad()
+def get_sentiment_scores(sentences, model, tokenizer):
+    """
+    Runs the model on sentences and returns a simplified -1, 0, +1 score per sentence.
+    """
     if not sentences:
         return []
+        
+    inputs = tokenizer(sentences, return_tensors="pt", padding=True, truncation=True, max_length=128)
+    outputs = model(**inputs)
+    predictions = torch.argmax(outputs.logits, dim=-1)
     
-    vectors = []
-    with torch.no_grad():
-        for sentence in sentences:
-            inputs = tokenizer(sentence, return_tensors="pt", padding=True, truncation=True, max_length=256)
-            logits = model(**inputs).logits
-            probabilities = torch.nn.functional.softmax(logits, dim=1).numpy()[0]
-            vectors.append(probabilities)
-    return vectors
-
-
-def aggregate_vectors_to_features(vectors, model, prefix=''):
-    """Aggregates sentence vectors into call-level features."""
-    if not vectors:
-        # Returning a dictionary with zero values for all emotions if no vectors
-        return {f"{prefix}{model.config.id2label[i]}_mean": 0 for i in range(model.config.num_labels)}
-
-    vectors_np = np.array(vectors)
-    # Calculating the mean probability for each emotion across all sentences
-    mean_sentiments = np.mean(vectors_np, axis=0)
+    score_map = {
+        'joy': 1, 'surprise': 1,
+        'anger': -1, 'fear': -1, 'sadness': -1,
+        'neutral': 0, 'shame': 0, 'disgust': 0
+    }
     
-    # Creating a feature dictionary
-    features = {f"{prefix}{model.config.id2label[i]}_mean": mean_sentiments[i] for i in range(len(mean_sentiments))}
-    return features
+    labels = [model.config.id2label[p.item()] for p in predictions]
+    scores = [score_map.get(label, 0) for label in labels]
+    return scores
 
-
-def get_stock_returns(ticker, earnings_date):
-    """Fetches stock prices and calculates 1-day and 5-day returns."""
-    earnings_date = pd.to_datetime(earnings_date)
-    start_date = earnings_date - timedelta(days=1)
-    end_date = earnings_date + timedelta(days=7)
+def aggregate_sentiment_features(scores):
+    """Aggregates a list of scores into features like mean score and positive ratio."""
+    if not scores:
+        return {'mean_score': 0, 'positive_ratio': 0, 'negative_ratio': 0}
     
-    stock_data = yf.download(ticker, start=start_date, end=end_date, progress=False)
+    total_sentences = len(scores)
+    positive_sentences = sum(1 for s in scores if s > 0)
+    negative_sentences = sum(1 for s in scores if s < 0)
     
-    if stock_data.empty: return None
+    return {
+        'mean_score': np.mean(scores),
+        'positive_ratio': positive_sentences / total_sentences,
+        'negative_ratio': negative_sentences / total_sentences
+    }
 
+def get_stock_returns(ticker, earnings_date_str):
+    """
+    More robustly fetches stock prices and calculates 1-day and 5-day returns.
+    """
     try:
-        price_on_earnings_date = stock_data.loc[stock_data.index.date == earnings_date.date()]['Adj Close'].iloc[0]
-    except IndexError:
-        try:
-            price_on_earnings_date = stock_data[stock_data.index > earnings_date]['Adj Close'].iloc[0]
-        except IndexError: return None
+        earnings_date = datetime.strptime(earnings_date_str, '%Y-%m-%d')
+        start_date = earnings_date - timedelta(days=1)
+        end_date = earnings_date + timedelta(days=10)
+        
+        stock_data = yf.download(ticker, start=start_date, end=end_date, auto_adjust=True, progress=False)
+        
+        if stock_data.empty:
+            return None
 
-    try:
-        price_1_day_after = stock_data[stock_data.index > earnings_date]['Adj Close'].iloc[0]
-        price_5_days_after = stock_data[stock_data.index > earnings_date]['Adj Close'].iloc[4]
+        price_on_date_series = stock_data.loc[stock_data.index >= earnings_date]
+        if price_on_date_series.empty: return None
+        # --- KEY CHANGE: Use 'Close' instead of 'Adj Close' ---
+        price_on_date = price_on_date_series.iloc[0]['Close']
         
-        return_1d = (price_1_day_after - price_on_earnings_date) / price_on_earnings_date
-        return_5d = (price_5_days_after - price_on_earnings_date) / price_on_earnings_date
-        return {'return_1d': return_1d, 'return_5d': return_5d}
-    except IndexError: return None
+        post_earnings_prices = stock_data.loc[stock_data.index > earnings_date]
+        if len(post_earnings_prices) < 5: return None
+
+        # --- KEY CHANGE: Use 'Close' instead of 'Adj Close' ---
+        price_1d_after = post_earnings_prices.iloc[0]['Close']
+        price_5d_after = post_earnings_prices.iloc[4]['Close']
+        # --- END OF CHANGE ---
+        
+        return {
+            'return_1d': (price_1d_after - price_on_date) / price_on_date,
+            'return_5d': (price_5d_after - price_on_date) / price_on_date
+        }
+    except Exception:
+        return None
+

finance/sample_transcripts_advanced.csv

@@ -0,0 +1,4 @@
+company_name,ticker,earnings_date,transcript
+"Apple Inc.","AAPL","2024-05-02","Good afternoon, everyone. We are thrilled to report another record-breaking quarter. Our iPhone sales continue to show incredible strength, and we are seeing fantastic momentum in our services division. The macro-economic environment presents some headwinds, but we are confident in our product pipeline. We are managing our supply chain effectively despite some challenges. The future looks very bright, and we are excited about the innovations we have in store. We believe we are well-positioned for strong growth ahead."
+"Microsoft Corp.","MSFT","2024-04-25","Thank you for joining us. This quarter's results demonstrate the strength of our cloud offerings. Azure's growth has accelerated, and we are seeing strong demand for our AI-powered services. We are navigating a complex global market, which has created some areas of softness, particularly in the PC market. However, our commercial bookings remain very strong. We are investing heavily in the future, and we see a long runway for growth. We are concerned about the regulatory landscape, but we are prepared to adapt."
+"NVIDIA Corp.","NVDA","2024-05-22","We have had a phenomenal quarter, driven by the unprecedented demand for our AI and data center platforms. The computational needs of the world are growing exponentially, and we are at the center of this transformation. We are facing some supply constraints, which is our biggest challenge. We are working tirelessly to increase our capacity. The gaming market has shown some weakness, but the data center segment is more than compensating for it. We are extremely optimistic about our future."