dashboard.py

"""import pandas as pd
import json
import matplotlib.pyplot as plt
from sklearn.linear_model import LinearRegression
import numpy as np

def analyze_data():
    # Load JSON data
    with open("session_data.json", "r") as file:
        data = json.load(file)

    # Extract data
    interactions = data["interactions"]
    timestamp = data.get("timestamp", "No Timestamp")

    # Parse interactions into a DataFrame for analysis
    interaction_data = []
    for interaction in interactions:
        sentiment = interaction["sentiment"]

        if isinstance(sentiment, list) and len(sentiment) > 0:
            sentiment_label = sentiment[0]["label"]
            sentiment_score = sentiment[0]["score"]
        elif isinstance(sentiment, dict):
            sentiment_label = sentiment.get("label", "No Sentiment")
            sentiment_score = sentiment.get("score", 0.0)
        else:
            sentiment_label = "No Sentiment"
            sentiment_score = 0.0

        sentiment_change = interaction["sentiment_change"]
        recommendations = interaction["product_recommendations"]
        objection = interaction["objection_handling"]

        interaction_data.append({
            "Transcription": interaction["transcription"],
            "Sentiment Label": sentiment_label,
            "Sentiment Score": sentiment_score,
            "Sentiment Change": sentiment_change,
            "Objection": objection[0] if objection else None,
            "Objection Response": objection[1] if len(objection) > 1 else None,
            "Recommendations": recommendations,
        })

    df = pd.DataFrame(interaction_data)

    # Convert list of recommendations to a string (hashable type)
    df['Recommendations_Str'] = df['Recommendations'].apply(lambda x: ', '.join([rec[0] for rec in x]) if isinstance(x, list) else str(x))
    df['Interaction Count'] = df.groupby('Recommendations_Str')['Recommendations_Str'].transform('count')
    df['CLV'] = df['Interaction Count'] * 10  # Example formula

    # Display data
    print(f"Data Timestamp: {timestamp}")
    print("\nCustomer Interaction Summary:")
    print(df)

    # Insights: Sentiment Trends (Pie chart)
    sentiment_counts = df["Sentiment Label"].value_counts()
    fig, ax = plt.subplots(figsize=(4, 4))
    sentiment_counts.plot(kind="pie", autopct="%1.1f%%", ax=ax)
    ax.set_ylabel("")
    plt.title("Sentiment Trends")
    plt.show()

    # Product Recommendations (Bar chart)
    all_recommendations = [rec[0] for recs in df["Recommendations"] for rec in recs]
    recommendation_counts = pd.Series(all_recommendations).value_counts()
    fig, ax = plt.subplots(figsize=(5, 3))
    recommendation_counts.plot(kind="bar", color="skyblue", ax=ax)
    ax.set_title("Top Products")
    plt.show()

    # Predictive Modeling for Sentiment Scores
    if len(df) > 1:
        X = np.arange(len(df)).reshape(-1, 1)
        y = df["Sentiment Score"].fillna(0).values

        model = LinearRegression()
        model.fit(X, y)

        future_steps = np.arange(len(df), len(df) + 5).reshape(-1, 1)
        predicted_scores = model.predict(future_steps)

        print("\nPredicted Sentiment Scores (Next 5 Interactions):")
        print(predicted_scores)

        # Visualization
        fig, ax = plt.subplots(figsize=(6, 4))
        ax.plot(range(len(df)), y, label="Actual Scores", marker="o")
        ax.plot(range(len(df), len(df) + 5), predicted_scores, label="Predicted Scores", linestyle="--", marker="o")
        ax.set_title("Sentiment Score Trends")
        ax.set_xlabel("Interaction Index")
        ax.set_ylabel("Sentiment Score")
        ax.legend()
        plt.show()

    # AI Recommendations for Sales Improvement
   # print("\nAI Recommendations for Sales Improvement:")
    #print("""
    #1. Address objections related to pricing and promotions.
    #2. Highlight the most recommended products to align with customer preferences.
    #3. Use sentiment trends to identify weak conversation points.
    #4. Optimize follow-ups based on predicted sentiment scores.
    #""")

    # Display Customer Lifetime Value (CLV)
    #print("\nCustomer Lifetime Value (CLV):")
    #print(df[['Recommendations_Str', 'Interaction Count', 'CLV']])

#if __name__ == "__main__":
    #analyze_data()"""
import pandas as pd
import json
import matplotlib.pyplot as plt
from sklearn.linear_model import LinearRegression
import numpy as np
from wordcloud import WordCloud

def analyze_data(session_data):
    # Extract data from the passed session data
    interactions = session_data["interactions"]
    timestamp = session_data.get("timestamp", "No Timestamp")

    # Parse interactions into a DataFrame
    interaction_data = []
    all_transcriptions = []
    for interaction in interactions:
        sentiment = interaction["sentiment"]

        if isinstance(sentiment, list) and len(sentiment) > 0:
            sentiment_label = sentiment[0]["label"]
            sentiment_score = sentiment[0]["score"]
        elif isinstance(sentiment, dict):
            sentiment_label = sentiment.get("label", "No Sentiment")
            sentiment_score = sentiment.get("score", 0.0)
        else:
            sentiment_label = "No Sentiment"
            sentiment_score = 0.0

        recommendations = interaction["product_recommendations"]
        objection = interaction["objection_handling"]
        transcription = interaction["transcription"]
        all_transcriptions.append(transcription)

        interaction_data.append({
    "Transcription": transcription,
    "Sentiment Label": sentiment_label,
    "Sentiment Score": sentiment_score,
    "Objection": objection.get("objection") if objection else None,
    "Objection Response": objection.get("response") if objection else None,
    "Recommendations": recommendations,
})


    df = pd.DataFrame(interaction_data)

    # Handle insufficient data for sentiment predictions
    if len(df) <= 1:
        sentiment_predictions = "Insufficient data for predictions"
    else:
        # Predictive Modeling for Sentiment Scores
        X = np.arange(len(df)).reshape(-1, 1)
        y = df["Sentiment Score"].fillna(0).values

        model = LinearRegression()
        model.fit(X, y)

        future_steps = np.arange(len(df), len(df) + 5).reshape(-1, 1)
        sentiment_predictions = model.predict(future_steps)

    # Generate Insights
    sentiment_counts = df["Sentiment Label"].value_counts()

    # Create a pie chart for sentiment trends
    fig_sentiment, ax = plt.subplots(figsize=(4, 4))
    sentiment_counts.plot(kind="pie", autopct="%1.1f%%", ax=ax)
    ax.set_ylabel("")
    plt.title("Sentiment Trends")

    # Create a bar chart for product recommendations
    all_recommendations = [rec[0] for recs in df["Recommendations"] for rec in recs]
    recommendation_counts = pd.Series(all_recommendations).value_counts()
    fig_recommendations, ax = plt.subplots(figsize=(5, 3))
    recommendation_counts.plot(kind="bar", color="skyblue", ax=ax)
    ax.set_title("Top Products")

    # Generate a Word Cloud for Transcriptions
    all_text = " ".join(all_transcriptions)
    wordcloud = WordCloud(background_color="white").generate(all_text)
    fig_wordcloud, ax = plt.subplots(figsize=(6, 4))
    ax.imshow(wordcloud, interpolation="bilinear")
    ax.axis("off")
    plt.title("Call Topics")

    # Actionable Recommendations
    actionable_recommendations = [
        "Focus on objection handling related to pricing concerns.",
        "Highlight top-performing products during calls.",
        "Leverage sentiment trends to adjust tone and messaging.",
        "Plan follow-ups based on predicted sentiment trends."
    ]

    # Return results for Streamlit display
    return {
        "timestamp": timestamp,
        "summary_table": df,
        "sentiment_chart": fig_sentiment,
        "recommendation_chart": fig_recommendations,
        "sentiment_predictions": sentiment_predictions,
        "wordcloud": fig_wordcloud,
        "actionable_recommendations": actionable_recommendations,
    }