Krivosheenkova / Sberbank-Russian-Housing-Market_docker

Flask application on Docker for predicting real estate prices on the Russian market.

Sberbank-Russian-Housing-Market_docker

Flask applications on Docker for predicting real estate prices on the Russian market.

Stack:

ML: sklearn, pandas, numpy, xgboost

API: Flask

Platform: Docker

Data: from kaggle - https://www.kaggle.com/c/sberbank-russian-housing-market

Task:

prediction of real estate prices, regression task

290 features in use. The model was fitted on 30471 observations.

Feature transformations: TimestampEncoding, LabelEncoding, SimpleImputer, MinMaxScaler

Model: XGBoostRegressor

Clone repo and build image

$ git clone https://github.com/Krivosheenkova/Sberbank-Russian-Housing-Market_docker.git
$ cd Sberbank-Russian-Housing-Market_docker
$ docker build -t sberbank-housing-market_docker_flask ./docker-flask-sberbank/

Run container

Here you need to create a directory locally and save the pre-trained model there (<your_local_path_to_pretrained_models> you need to replace with the path to this directory)

$ docker run -d -p 4140:4140 -v <your_local_path_to_pretrained_models>:/app/app/models sberbank-housing-market_docker_flask

Run get_predictions.py

$ python get_predictions.py <path_to_csv_file> --outfile predictions.csv