house-prices-prediction-LGBM

Description

This repo has been developed for the Istanbul Data Science Bootcamp, organized in cooperation with İBB & Kodluyoruz. Prediction for house prices was developed using the Kaggle House Prices - Advanced Regression Techniques competition dataset.

Data

The dataset is available at Kaggle.

Goal

The goal of this project is to predict the price of a house in Ames using the features provided by the dataset.

Features

The dataset contains the following features:

OverallQual: Overall quality of the house
GrLivArea: Above grade (ground) living area square feet
GarageCars: Number of garage cars
TotalBsmtSF: Total square feet of basement area
FullBath: Number of full baths
YearBuilt: Year house was built
TotRmsAbvGrd: Total number of rooms above grade (excluding bathrooms and closets)
Fireplaces: Number of fireplaces
BedroomAbvGr: Number of bedrooms above grade
GarageYrBlt: Year garage was built
LowQualFinSF: Lowest quality finished square feet
LotFrontage: Lot frontage square feet
MasVnrArea: Masonry veneer square feet
WoodDeckSF: Square feet of wood deck area
OpenPorchSF: Open porch square feet
EnclosedPorch: Enclosed porch square feet
3SsnPorch: Three season porch square feet
ScreenPorch: Screen porch square feet
PoolArea: Pool square feet
MiscVal: Miscellaneous value
MoSold: Month house was sold
YrSold: Year house was sold
SalePrice: Sale price

Usage

# clone the repo
git clone https://github.com/uzunb/house-prices-prediction-LGBM.git

# change to the repo directory
cd house-prices-prediction-LGBM

# if virtualenv is not installed, install it
#pip install virtualenv

# create a virtualenv
virtualenv -p python3 venv

# activate virtualenv for LINUX or MACOS
source venv/bin/activate

# # activate virtualenv for WINDOWS
# venv\Scripts\activate.ps1
#     # throubleshooting for activation error in windows
#     Set-ExecutionPolicy RemoteSigned -Scope CurrentUser

# install dependencies
pip install -r requirements.txt

# run the script
streamlit run main.py

Model Development

Model

The model is based on a LightGBM algorithm.

Training

import lightgbm as lgb

model = lgb.LGBMRegressor(max_depth=3, 
                    n_estimators = 100, 
                    learning_rate = 0.2,
                    min_child_samples = 10)
model.fit(x_train, y_train)

Grid Search Cross Validation is used for hyper parameters of the model.

from sklearn.model_selection import GridSearchCV

params = [{"max_depth":[3, 5], 
            "n_estimators" : [50, 100], 
            "learning_rate" : [0.1, 0.2],
            "min_child_samples" : [20, 10]}]

gs_knn = GridSearchCV(model,
                      param_grid=params,
                      cv=5)

gs_knn.fit(x_train, y_train)
gs_knn.score(x_train, y_train)

pred_y_train = model.predict(x_train)
pred_y_test = model.predict(x_test)

r2_train = metrics.r2_score(y_train, pred_y_train)
r2_test = metrics.r2_score(y_test, pred_y_test)

msle_train =metrics.mean_squared_log_error(y_train, pred_y_train)
msle_test =metrics.mean_squared_log_error(y_test, pred_y_test)

print(f"Train r2 = {r2_train:.2f} \nTest r2 = {r2_test:.2f}")
print(f"Train msle = {msle_train:.2f} \nTest msle = {msle_test:.2f}")

print(gs_knn.best_params_)

Evaluation

from sklearn.metrics import mean_squared_error
from sklearn.metrics import mean_absolute_error
from sklearn.metrics import r2_score
from sklearn.metrics import explained_variance_score
from sklearn.metrics import mean_squared_log_error

y_pred = model.predict(x_test)
print('Mean Absolute Error:', mean_absolute_error(y_test, y_pred))
print('Mean Squared Error:', mean_squared_error(y_test, y_pred))
print('Root Mean Squared Error:', np.sqrt(mean_squared_error(y_test, y_pred)))
print('Mean Squared Log Error:', mean_squared_log_error(y_test, y_pred))
print('Explained Variance Score:', explained_variance_score(y_test, y_pred))
print('R2 Score:', r2_score(y_test, y_pred))

Deployment

Simple model distribution is made using Streamlit.

import streamlit as st

st.title("House Prices Prediction")
st.write("This is a simple model for house prices prediction.")

st.sidebar.title("Model Parameters")

variables = droppedDf["Alley"].drop_duplicates().to_list()
inputDict["Alley"] = st.sidebar.selectbox("Alley", options=variables)

inputDict["LotFrontage"] = st.sidebar.slider("LotFrontage", ceil(droppedDf["LotFrontage"].min()), 
floor(droppedDf["LotFrontage"].max()), int(droppedDf["LotFrontage"].mean()))

Results

The model is trained on the dataset and tested on the test dataset. The results are shown demo with Streamlit below:

Contributions

Batuhan UZUN - Github - LinkedIn
Dursun Tunahan BİLGİN - Github - LinkedIn
Anıl DÖNMEZ - Github - LinkedIn
Hazal SEZGİN - Github - LinkedIn
Müşerref ÖZKAN - Github - LinkedIn
Hanife YAMAN - Github - LinkedIn
Üftade Bengi EROLÇAY - Github - LinkedIn
Yiğit YILMAZ - Github - LinkedIn
Selin ÇILDAM - Github - LinkedIn

uzunb / house-prices-prediction-LGBM