Project Work: Customer Behavior Prediction

The project involves the development of a web app for predicting the future annual expenses of customers using an e-commerce service.

The training and testing of the ML model were carried out using a Kubeflow pipeline.

In particular, the code within the current directory "CustomerBehaviorPrediction" is divided into four subdirectories:

• /app
• /cluster
• /pipeline
• /NGINX_example

App

Within the /app directory, the following items are present:

• app.py: code for the web application
• Dockerfile: Dockerfile for creating the application container
• k8s_customer_behavior_deployment.yaml: file for configuring the pods related to the application in the cluster (Deployment & Service)
• requirements.txt: Python libraries required for running the application, which will be included in the container
• /model: directory containing the scaler and model used by the application
• deployment.md: file explaining how to deploy and run the application on the cluster
• docker_hub.sh: script for creating the application container and pushing the image to Docker Hub

Cluster

Within the /cluster directory, the following items are present:

• cluster_config.yaml: Kubernetes cluster configuration file
• metric-server.yaml: metric-server configuration file
• cluster.md: file explaining how to set up the cluster and install the metric-server

Pipeline

Within the /pipeline directory, the following items are present:

• /load_data: Component responsible for loading the dataset for training
• /preprocess_data: Component responsible for data preprocessing
• /linear_regression: Component responsible for training and testing the Linear Regression model
• /xgboost_regressor: Component responsible for training and testing the XGBoost Regressor model
• /neural_regression: Component responsible for training and testing the MLP model
• customer_pipeline.py: file for generating the kubeflow pipeline YAML file (also contains the definition of the show_result and evaluate_best_model components)
• customer_pipeline.yaml: kubeflow pipeline YAML file to be inserted into kubeflow
• docker_hub.sh: script for creating containers for various components and pushing images to Docker Hub
• pipeline.md: file explaining how to install Kubeflow on the Kubernetes cluster and import the pipeline

NGINX_example

Within the /NGINX_example directory, the following items are present:

• IngressController.yaml: configuration file for the Ingress Controller for the Web App Firewall
• VirtualServer.yaml: configuration file for the virtual server of the Web App Firewall