Prediction Model

Prediction model for Mobi.lo apps by Mari Bangkit. Hosted in Azure ML as a web service.

About the Model

The exported model uses Lasso Regression to predict the car purchase amount a customer would buy based on their profile, such as gender, annual salary, and more. The main aim is to provide a personalized car recommendation for customers.

The dataset used is accessible here. It is originally come from Car Purchase Price (beginner dataset), but has been modified to suit the Indonesian currency (Rp).

Publication

Research paper: https://ejurnal.stmik-budidarma.ac.id/index.php/jurikom/article/view/5021

Deployment

Prerequisites

Azure Machine Learning workspace resource.
Azure CLI and the ml extension. Each on v2 or more.
Python v3 or more.
Bash/Fish/ZSH or any UNIX-like shell.

Local Deployment

Clone the repo.

git clone https://github.com/mari-bangkit/prediction-model.git
cd prediction-model

az login
az account set --subscription <subscription ID>
az configure --defaults group=<resource group> workspace=<Azure Machine Learning workspace name>

Run local deployment script.
```
./deploy_local.sh
```
The endpoint URL is provided at the FINAL STATUS section.

Test the API.

Bash/ZSH shell:

export AZURE_ML_URL_LOCAL=<local endpoint url>
python test_request_local.py

Fish shell:

set -x AZURE_ML_URL_LOCAL <local endpoint url>
python test_request_local.py

Azure Deployment

Run Azure deployment script.
```
./deploy_azure.sh
```
Open your workspace in Azure ML Studio to get the API URL and key.

Test the API.

Bash/ZSH shell:

export AZURE_ML_URL=<endpoint url>
export AZURE_ML_API_KEY=<endpoint key>
python test_request_cloud.py

Fish shell:

set -x AZURE_ML_URL <endpoint url>
set -x AZURE_ML_API_KEY <endpoint key>
python test_request_cloud.py