User Conversion API

This project provides a containerized Flask-based API for predicting TMall's user conversion using a pre-trained Random Forest model. The API processes JSON input and returns a prediction along with the confidence level.

---

Team Members

Lingling Deng, Chih-Yun Chang, Yao Chen, Lixing Chen

Link To Final Presentation

Presentation Link

Link To Final Report

Report Link

---

Prediction Endpoint from GCP

URL: Open your browser or API testing tool (e.g., Postman) and send a POST request to https://user-conversion-api-999732714550.us-east1.run.app/predict

Method: POST

Headers:

{
  "Content-Type": "application/json"
}

Use Following CURL Commands In Terminals

Sample Input 1:

curl -X POST \
  https://user-conversion-api-999732714550.us-east1.run.app/predict \
  -H "Content-Type: application/json" \
  -d '{
    "Income": 39192,
    "CampaignChannel": 0,
    "CampaignType": 0,
    "AdSpend": 429.07921736744015,
    "ClickThroughRate": 0.0728897285108404,
    "ConversionRate": 0.09538896175545068,
    "WebsiteVisits": 44,
    "PagesPerVisit": 6.95369491204816,
    "TimeOnSite": 2.2520393830350898,
    "SocialShares": 33,
    "EmailOpens": 0,
    "EmailClicks": 1,
    "PreviousPurchases": 1,
    "LoyaltyPoints": 4474
  }'

Expected Output 1:

{
    "confidence": 0.88,
    "prediction": 0
}

Sample Input 2:

curl -X POST \
  https://user-conversion-api-999732714550.us-east1.run.app/predict \
  -H "Content-Type: application/json" \
  -d '{
    "Income": 55972.0,
    "AdSpend": 7254.022157320001,
    "ClickThroughRate": 0.083025001631625,
    "ConversionRate": 0.015726698375027437,
    "WebsiteVisits": 38,
    "PagesPerVisit": 5.447257523959778,
    "TimeOnSite": 6.134174721060454,
    "SocialShares": 79,
    "EmailOpens": 2,
    "EmailClicks": 8,
    "PreviousPurchases": 1,
    "LoyaltyPoints": 612
  }'

Expected Output 2:

{
    "confidence": 0.82,
    "prediction": 0
}

Sample Input 3:

curl -X POST \
  https://user-conversion-api-999732714550.us-east1.run.app/predict \
  -H "Content-Type: application/json" \
  -d '{
    "Income": 109779.0,
    "AdSpend": 8383.984491524046,
    "ClickThroughRate": 0.28263938781591186,
    "ConversionRate": 0.17492572533137893,
    "WebsiteVisits": 6,
    "PagesPerVisit": 2.384030881810251,
    "TimeOnSite": 12.561754154488309,
    "SocialShares": 97,
    "EmailOpens": 14,
    "EmailClicks": 2,
    "PreviousPurchases": 3,
    "LoyaltyPoints": 467
  }'

Expected Output 3:

{
    "confidence": 0.79,
    "prediction": 1
}

Sample Input 4:

curl -X POST \
  https://user-conversion-api-999732714550.us-east1.run.app/predict \
  -H "Content-Type: application/json" \
  -d '{
    "Income": 136912,
    "AdSpend": 6497.870068417766,
    "ClickThroughRate": 0.04391851073538301,
    "ConversionRate": 0.08803141207288108,
    "WebsiteVisits": 0,
    "PagesPerVisit": 2.399016527783845,
    "TimeOnSite": 7.3968025807960585,
    "SocialShares": 19,
    "EmailOpens": 6,
    "EmailClicks": 9,
    "PreviousPurchases": 4,
    "LoyaltyPoints": 688
  }'

Expected Output 4:

{
    "confidence": 0.94,
    "prediction": 1
}

---

Requirements

To run locally:

• Python 3.9+
• Docker

---

Setup

Running Locally

1. Clone the repository:

git clone https://github.com/chrischenlixing/CS6220FinalProject.git
cd CS6220FinalProject

2. Install dependencies:

pip install -r requirements.txt

3. Generate the model file:

python user_conversion.py

4. Run the Flask application:

python app.py

5. Access the API:: Open your browser or API testing tool (e.g., Postman) and send a POST request to http://127.0.0.1:8080/predict

Running with Docker

1. Build the Docker image::

docker build -t user-conversion-api .

2. Run the Docker container::

docker run -p 8080:8080 user-conversion-api

---