The problem statement is to predict the sale price of a house, given the features of the house. The features are the columns in the dataset, and the target variable is the SalePrice column. The problem is a regression problem, as the target variable is continuous.
The data description is available on Kaggle, and you can find it here. The data description contains a detailed description of each column in the dataset. The data description is very useful, as it provides a detailed description of each column in the dataset. It also provides information about the missing values in the dataset.
Python Code:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
data = pd.read_csv('train.csv')
# Displaying all the columns in the dataset
pd.set_option('display.max_columns', None)
print(data.head())
# Making a list of columns with missing values missing_values = [col for col in data.columns if data[col].isnull().any()]
Output
# Dropping the columns with more than 15% missing values data.drop(['Alley', 'FireplaceQu', 'PoolQC', 'Fence', 'MiscFeature'], axis=1, inplace=True)
# Filling the missing values in the remaining columns with the most frequent value new_missing_values = [col for col in data.columns if data[col].isnull().any()] for col in new_missing_values: if data[col].dtype == 'O': data[col].fillna(data[col].mode()[0], inplace=True) else: data[col].fillna(data[col].median(), inplace=True)
continuous_features = [col for col in data.columns if data[col].dtype != 'O'] for col in continuous_features: data_copy = data.copy() if 0 in data_copy[col].unique(): pass else: data_copy[col] = np.log(data_copy[col]) data_copy['SalePrice'] = np.log(data_copy['SalePrice']) plt.scatter(data_copy[col], data_copy['SalePrice']) plt.xlabel(col) plt.ylabel('SalePrice')
Output
plt.figure(figsize=(10, 15)) # Plotting the heatmap with respect to the correlation of the features with the target variable 'SalePrice' sns.heatmap(data.corr()[['SalePrice']].sort_values(by='SalePrice', ascending=False), annot=True, cmap='viridis')
Output
# Regression plot between the target variable and the most correlated variables who have a correlation greater than 0.5 for col in data.corr()[data.corr()['SalePrice'] > 0.5].index: if col == 'SalePrice': pass else: sns.regplot(x=data[col], y=data['SalePrice']) plt.show()
Output
# Encoding the categorical variables to the numeric datatype from sklearn.preprocessing import LabelEncoder for col in data.columns: if data[col].dtype == 'O': label_encoder = LabelEncoder() data[col] = label_encoder.fit_transform(data[col])
# Let's use Ridge Regression to build the model from sklearn.linear_model import Ridge from sklearn.model_selection import cross_val_score from sklearn.metrics import mean_squared_error
# create X and y variables from Features and target variable X = data[['OverallQual', 'GrLivArea', 'GarageCars', 'GarageArea', 'TotalBsmtSF', '1stFlrSF', 'FullBath', 'TotRmsAbvGrd', 'YearBuilt', 'YearRemodAdd']] y = data['SalePrice']
# Function to perform ridge regression def ridge_regression(alpha, data): ridge = Ridge(alpha=alpha) ridge.fit(X, y) scores = cross_val_score(ridge, X, y, scoring='neg_mean_squared_error', cv=5) rmse = np.sqrt(-scores) return rmse
# Finding the best value of hyper-parameter of Ridge - alpha alpha = [0.001, 0.01, 0.1, 1, 10, 100, 1000] for i in alpha: print('Alpha: ', i) print('Mean RMSE: ', ridge_regression(i, data).mean()) print('Standard Deviation: ', ridge_regression(i, data).std()) print()
Output
# I will be using the ridge regression model with alpha = 100 ridge = Ridge(alpha=100) ridge.fit(X, y)
# Loading the test data test_data = pd.read_csv('test.csv')
# Doing all the changes that were done in the training data test_data.drop(['Alley', 'FireplaceQu', 'PoolQC', 'Fence', 'MiscFeature'], axis=1, inplace=True) new_missing_values = [col for col in test_data.columns if test_data[col].isnull().any()] for col in new_missing_values: if test_data[col].dtype == 'O': test_data[col].fillna(test_data[col].mode()[0], inplace=True) else: test_data[col].fillna(test_data[col].median(), inplace=True)
# Encoding the categorical variables of test data for col in test_data.columns: if test_data[col].dtype == 'O': label_encoder = LabelEncoder() test_data[col] = label_encoder.fit_transform(test_data[col])
# Selecting Features X_test = test_data[['OverallQual', 'GrLivArea', 'GarageCars', 'GarageArea', 'TotalBsmtSF', '1stFlrSF', 'FullBath', 'TotRmsAbvGrd', 'YearBuilt', 'YearRemodAdd']] # Predicting the target variable y_pred = ridge.predict(X_test)
# Saving the predictions in a csv file output = pd.DataFrame({'Id': test_data.Id, 'SalePrice': y_pred}) output.to_csv('my_submission.csv', index=False)
In this article, we have chosen the Ames housing dataset as the price prediction model, understand the problem statement, and perform Exploratory Data Analysis. We have also performed missing value imputation and encoding of categorical variables on the train and test data sets. Then we applied Ridge regression with different alpha values and found the best value of alpha to minimize the RMSE.
Key Takeaways:
The media shown in this article is not owned by Analytics Vidhya and is used at the Authorâs discretion.