zemerov / end2end_project

Project for end to end simple ML project

Simple flask app for price prediction

This is a flask server app for predicting real estate price using machine learning. Project for end to end simple ML project in GSOM.

Model

The predicting model is a gradient boosting using catboost lib. It has the best fit among other methods: linear regression and random forest.

Used hyperparameters for this model

num_trees=500
random_seed=42
depth=5

Other hyperparameters are set to default values.

Setup

You can set up a server either using virtual env or docker container. Just follow the instruction.

Using virtual env

First step: create virtual env

python -m venv venv

Then activate this

source venv/bin/activate

Install requirements

pip install -r requirements.txt

Start the app on port 80 (default value)

python src/main.py

If you want to use custom argument try this:

python src/main.py --port <port> --model-path <path_to_mode>

Using docker

You can use existing Dockerfile in this repo. It has all necessary commands to run this server. It uses 80 as default.

Build docker

sudo docker build . -t server

Start docker and pass 80 port

sudo docker run -p 80:80 -t server

Congratulations! Your application is running on port 80

Request

Consider your flask app is running on a remote host on a specific port.

Request example:

curl "http://<remote_host_ip>:<port>/predict_price?rooms=3&area=32"

List of supported arguments:

Name	Type
floor	int
category_type	int
open_plan	bool
rooms	int
studio	bool
area	float
kitchen_area	float
living_area	float
agent_fee	float
renovation	bool
offer_type	int

Open port on a remote machine

sudo ufw allow 80