ML Project: Sber Auto Subscription

Table of Contents

1. Introduction
2. Project Structure
3. Installation
4. Usage

Introduction

This project involves developing a machine learning model for a car subscription service. The main task is to predict whether a user will perform a target action on the company's website. Key aspects of the project include:

1. Working with Google Analytics data to analyze user behavior on the website.
2. Developing a predictive model to determine if a user will perform a target action (e.g., submitting an application or requesting a call).
3. The goal is to achieve a ROC-AUC score of approximately 0.65.
4. Creating a service that takes input features such as UTM parameters, device information, and geo data, and outputs a binary prediction (0 or 1) indicating whether the user is likely to perform a target action.
5. The project requires data preprocessing, feature engineering, model selection, and API implementation.

Project Structure

• data/skillbox_diploma_main_dataset_sberautopodpiska # Contains two datasets: information about sessions and information about hits
• notebooks/ # Jupyter notebooks for data preprocessing, EDA, feature engineering, and modeling
• model/ # Directory for model files:
  • pipeline.py # Steps for model creation
  • model.pkl # Final trained model
• main.py # FastAPI app configuration
• requirements.txt # Python dependencies
• README.md Project documentation

Installation

1. Clone the repository

   git@github.com:serverdaun/ml_project_sber.git
   cd ml_project_sber

2. Create and activate a virtual environment

   python3 -m venv venv
   source venv/bin/activate

3. Install the required dependencies

   pip install -r requirements.txt

Usage

1. Start the FastAPI application to serve the model and make predictions through a REST API.

   uvicorn main:app --reload

2. Once the server is running, you can use tools like 'curl' or Postman to make requests to the API. Below is the list of the calls.

• GET
  • 'http://127.0.0.1:8000/status' to get the status of the app
  • 'http://127.0.0.1:8000/version' to get model metadata
• POST
  • 'http://127.0.0.1:8000/predict' to make predictions with json format data in body

3. Example data for API

   {
   "session_id": "9055434745589932991.1637753792.1637753792",
   "client_id": "2108382700.1637753791",
   "visit_date": "2021-11-24",
   "visit_time": "14:36:32",
   "visit_nuber": 1,
   "utm_source": "ZpYIoDJMcFzVoPFsHGJL",
   "utm_medium": "banner",
   "utm_campaign": "LEoPHuyFvzoNfnzGgfcd",
   "utm_adcontent": "vCIpmpaGBnIQhyYNkXqp",
   "utm_keyword": "puhZPIYqKXeFPaUviSjo",
   "device_category": "mobile",
   "device_os": "Android",
   "device_brand": "Huawei",
   "device_model": "example_model",
   "device_screen_resolution": "360x720",
   "device_browser": "Chrome",
   "geo_country": "Russia",
   "geo_city": "Zlatoust"
   }