Upload 3 files

README.md

@@ -10,3 +10,36 @@ pinned: false
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+
+# AI Ad Server Optimizer
+
+## Overview
+The AI Ad Server Optimizer is a Python-based application designed to enhance ad performance through data-driven insights. It utilizes machine learning models to cluster user behavior and provides actionable insights to optimize ad targeting and engagement.
+
+## Features
+- **Cluster Prediction**: Predicts user behavior clusters based on session data.
+- **Ad Performance Analytics**: Summarizes key ad performance metrics like Click-Through Rate (CTR), Conversion Rate, and Bounce Rate.
+
+## Technology Stack
+- Python 3.8+
+- Pandas for data manipulation
+- Scikit-learn for machine learning tasks
+- Gradio for creating interactive web interfaces
+
+## Installation
+Ensure you have Python 3.8 or higher installed. Clone the repository and install the required dependencies:
+```bash
+git clone https://github.com/trilogy-group/Skyvera-AI.git
+cd scripts/AIAdServerOptimizer
+pip install -r requirements.txt
+```
+
+## Usage
+Run the application with:
+```bash
+python scripts/AIAdServerOptimizer/app.py
+```
+Navigate to the provided local URL to access the interactive web interface.
+
+## Contributing
+Contributions are welcome! Please fork the repository and submit pull requests with your enhancements. Ensure you follow the existing code style

app.py

@@ -0,0 +1,162 @@
+import pandas as pd
+from sklearn.cluster import KMeans
+from sklearn.preprocessing import StandardScaler, OneHotEncoder
+from sklearn.compose import ColumnTransformer
+from sklearn.pipeline import Pipeline
+import logging
+import gradio as gr
+
+# Configure logging
+logging.basicConfig(level=logging.DEBUG, format='%(asctime)s - %(levelname)s - %(message)s')
+
+# Expanded sample data
+data = pd.DataFrame({
+    'User ID': [1, 2, 3, 4, 5],
+    'Session Duration': [300, 450, 200, 600, 350],
+    'Pages Visited': [5, 8, 3, 12, 7],
+    'Ads Clicked': [2, 1, 0, 3, 2],
+    'User Interests': ['technology', 'sports', 'technology', 'arts', 'sports'],
+    'Engagement Score': [0.8, 0.5, 0.3, 0.9, 0.7],
+    'Device Type': ['mobile', 'desktop', 'mobile', 'tablet', 'desktop'],
+    'Time of Day': ['morning', 'afternoon', 'evening', 'morning', 'afternoon'],
+    'Time Spent per Page': [30, 25, 45, 20, 50],
+    'Click Through Rate': [0.1, 0.2, 0.05, 0.3, 0.15],
+    'Conversion Rate': [0.05, 0.1, 0, 0.2, 0.1],
+    'Frequency of Visits': [10, 20, 5, 15, 10],
+    'Bounce Rate': [0.2, 0.1, 0.5, 0.05, 0.3]
+})
+
+logging.info("Sample data prepared.")
+
+# Updated preprocessing
+preprocessor = ColumnTransformer(
+    transformers=[
+        ('num', StandardScaler(), ['Session Duration', 'Pages Visited', 'Ads Clicked', 'Engagement Score', 
+                                   'Time Spent per Page', 'Click Through Rate', 'Conversion Rate', 
+                                   'Frequency of Visits', 'Bounce Rate']),
+        ('cat', OneHotEncoder(), ['User Interests', 'Device Type', 'Time of Day'])
+    ])
+
+logging.info("Preprocessor setup complete.")
+
+# Clustering
+kmeans = KMeans(n_clusters=3, random_state=42)
+logging.info("KMeans clustering configured.")
+
+# Define the pipeline
+pipeline = Pipeline([
+    ('preprocessor', preprocessor),
+    ('cluster', kmeans)
+])
+
+logging.info("Pipeline created.")
+
+# Fit the pipeline to the data
+pipeline.fit(data)
+
+def generate_insights(cluster_characteristics):
+    # Example insights based on hypothetical thresholds
+    insights = []
+    if cluster_characteristics['Engagement Score'] > 0.7 and cluster_characteristics['Conversion Rate'] < 0.1:
+        insights.append("High engagement but low conversion: Consider optimizing the checkout process or providing targeted offers.")
+    if cluster_characteristics['Click Through Rate'] > 0.2:
+        insights.append("High click-through rate: Users are interacting well with ads. Increase ad relevance to boost conversions.")
+    if cluster_characteristics['Bounce Rate'] > 0.3:
+        insights.append("High bounce rate: Review landing page design and content relevance to improve user retention.")
+    return " ".join(insights)
+
+def predict_cluster(session_duration, pages_visited, ads_clicked, engagement_score, user_interests, device_type, time_of_day, time_spent_per_page, click_through_rate, conversion_rate, frequency_of_visits, bounce_rate):
+    input_df = pd.DataFrame({
+        'Session Duration': [session_duration],
+        'Pages Visited': [pages_visited],
+        'Ads Clicked': [ads_clicked],
+        'Engagement Score': [engagement_score],
+        'User Interests': [user_interests],
+        'Device Type': [device_type],
+        'Time of Day': [time_of_day],
+        'Time Spent per Page': [time_spent_per_page],
+        'Click Through Rate': [click_through_rate],
+        'Conversion Rate': [conversion_rate],
+        'Frequency of Visits': [frequency_of_visits],
+        'Bounce Rate': [bounce_rate]
+    })
+    cluster = pipeline.predict(input_df)[0]
+    centroids = pipeline.named_steps['cluster'].cluster_centers_
+    cluster_characteristics = centroids[cluster]
+
+    # Decode features for insights
+    num_features = ['Session Duration', 'Pages Visited', 'Ads Clicked', 'Engagement Score', 'Time Spent per Page', 'Click Through Rate', 'Conversion Rate', 'Frequency of Visits', 'Bounce Rate']
+    scaled_features = cluster_characteristics[:9]
+    original_num_values = pipeline.named_steps['preprocessor'].named_transformers_['num'].inverse_transform([scaled_features])[0]
+    cat_features = ['User Interests', 'Device Type', 'Time of Day']
+    encoded_features = cluster_characteristics[9:]
+    original_cat_values = pipeline.named_steps['preprocessor'].named_transformers_['cat'].inverse_transform([encoded_features])[0]
+
+    # Combine numerical and categorical features into a dictionary
+    cluster_characteristics = dict(zip(num_features, original_num_values))
+    cluster_characteristics.update(dict(zip(cat_features, original_cat_values)))
+
+    # Generate actionable insights
+    insights = generate_insights(cluster_characteristics)
+
+    return f"Predicted Cluster: {cluster}\nCharacteristics: {cluster_characteristics}\nActionable Insights: {insights}"
+
+def ad_performance_analytics():
+    # Calculate average metrics
+    avg_ctr = data['Click Through Rate'].mean()
+    avg_conversion_rate = data['Conversion Rate'].mean()
+    avg_bounce_rate = data['Bounce Rate'].mean()
+    
+    # Prepare the analytics report
+    report = f"Average Click Through Rate: {avg_ctr:.2%}\n"
+    report += f"Average Conversion Rate: {avg_conversion_rate:.2%}\n"
+    report += f"Average Bounce Rate: {avg_bounce_rate:.2%}"
+    
+    return report
+
+with gr.Blocks() as demo:
+    with gr.Tab("Cluster Prediction"):
+        with gr.Row():
+            gr.Markdown("**This form allows you to input user session data to predict which cluster the user belongs to and provides actionable insights based on their behavior.**")
+            session_duration = gr.Number(label="Session Duration", value=300)  # Set initial value
+            pages_visited = gr.Number(label="Pages Visited", value=5)  # Set initial value
+            ads_clicked = gr.Number(label="Ads Clicked", value=2)  # Set initial value
+            engagement_score = gr.Slider(0, 1, label="Engagement Score", value=0.5)  # Set initial value
+            user_interests = gr.Dropdown(['technology', 'sports', 'arts'], label="User Interests", value='technology')  # Set initial value
+            device_type = gr.Radio(['mobile', 'desktop', 'tablet'], label="Device Type", value='mobile')  # Set initial value
+            time_of_day = gr.Radio(['morning', 'afternoon', 'evening'], label="Time of Day", value='morning')  # Set initial value
+            time_spent_per_page = gr.Number(label="Time Spent per Page", value=30)  # Set initial value
+            click_through_rate = gr.Slider(0, 1, step=0.01, label="Click Through Rate", value=0.1)  # Set initial value
+            conversion_rate = gr.Slider(0, 1, step=0.01, label="Conversion Rate", value=0.05)  # Set initial value
+            frequency_of_visits = gr.Number(label="Frequency of Visits", value=10)  # Set initial value
+            bounce_rate = gr.Slider(0, 1, step=0.01, label="Bounce Rate", value=0.2)  # Set initial value
+        predict_button = gr.Button("Predict")
+        output_textbox = gr.Textbox(label="Prediction Output")
+        predict_button.click(
+            predict_cluster,
+            inputs=[
+                session_duration, pages_visited, ads_clicked, engagement_score, user_interests, device_type,
+                time_of_day, time_spent_per_page, click_through_rate, conversion_rate, frequency_of_visits, bounce_rate
+            ],
+            outputs=output_textbox
+        )
+
+    with gr.Tab("Ad Performance Analytics"):
+        gr.Markdown("""
+        **This form provides a summary of key performance metrics for ads.**
+        
+        - **Average Click-Through Rate (CTR):** Measures the percentage of ad views that result in clicks. Higher values indicate more effective ad engagement.
+        - **Average Conversion Rate:** Indicates the percentage of clicks that convert into actions, such as purchases or sign-ups. This metric helps assess the effectiveness of ad targeting and the overall conversion potential.
+        - **Average Bounce Rate:** Reflects the percentage of single-page visits. Lower bounce rates suggest that the landing pages are relevant to the visitors' interests.
+        
+        Understanding these metrics can help optimize ad strategies and improve overall campaign performance.
+        """)
+        analytics_button = gr.Button("Analyze Ad Performance")
+        analytics_output = gr.Textbox(label="Analytics Output")
+        analytics_button.click(
+            ad_performance_analytics,
+            outputs=analytics_output
+        )
+
+demo.launch()
+

requirements.txt

@@ -0,0 +1,7 @@
+pandas
+sklearn.cluster
+sklearn.preprocessing
+sklearn.compose
+sklearn.pipeline
+logging
+gradio