Predicting of Realty Prices

Introduction

This library implements ML model which is used to predict prices of realty in Moscow Region. The model is trained on Sberbank Russian Housing Market dataset

Usage

Train script

Trains and saves the Model

Template

[~]$ python train.py <data_path> <save_path>

Input parameters

positional arguments:

• data_path - Path of train data
• save_path - Path to save trained model

options:

• -h, --help - show this help message and exit
• --objective OBJECTIVE - Train objective
• --n_estimators N_ESTIMATORS - Number of gradient boosted trees
• --max_depth MAX_DEPTH - Maximum tree depth for base learners
• --eta ETA - Boosting learning rate
• --subsample SUBSAMPLE - Subsample ratio of the training instance
• --colsample_bytree COLSAMPLE_BYTREE - Subsample ratio of columns when constructing each tree
• --reg_lambda REG_LAMBDA - L2 regularization term on weights
• --random_state RANDOM_STATE - Random number seed
• --early_stopping_rounds EARLY_STOPPING_ROUNDS - Activates early stopping

Example

[~]$ ls
[~]$ python train.py train.csv .
[~]$ ls
model.json

Predict script

Predicts the prices and outputs it

Template

[~]$ python predict.py <data_path> <model_path> <save_path>

Input parameters

positional arguments:

• data_path - Path of data to predict
• model_path - Path of trained model
• save_path - Path of saved predictions

Example

[~]$ python predict.py <data_path> <model_path> <save_path>
[~]$ ls <save_path>
predictions.npy

Setup

Check out setup guide here