Abstract

House price valuation is critical when it comes to real estate selections. Mortgage lenders and house buyers and sellers utilise valuation estimations to gauge risk and the reasonableness of an asking price. In this thesis, models for predicting changes in property prices are examined, as well as the factors that drive those changes.

In real estate listings, a set list of numerical characteristics is used to describe the property. Spatial data, or location information, is a very important part of predicting home prices because the same house can sell for very different amounts in different places. We will investigate how each of these types of data affects how well we can predict home prices. This work shows how each type of data helps models that predict prices. This thesis makes a big contribution to making a software artefact by showing how a software implementation can be used to improve an algorithm for predicting house prices.

The housing market could use a better mechanism for projecting house values. The housing price projection can help a house seller, buyer, or real estate broker make more educated decisions. The regression method can still be beneficial because of its speed and ability to give generally accurate predictions. Regarding the prediction of property prices, we conclude that machine learning is a potential, alternative technique.

We have used the California House Price Prediction data for the current study. The purpose of this study is to create a model of home prices by making use of the information that has been provided to produce predictions about the median house value in the state of California. The corresponding dataset for the project includes 20,640 municipalities in total.

In the current study, to predict the median house price, appropriate regression models were applied which include appropriate parametric regression like simple linear regression model, ridge, lasso regression model as well as other regression models like Decision Tree Regression and Random Forest Regression Model. The Random Forest model proved to be the most appropriate model giving the highest value of the R-square and the minimum Root Mean Square Error Value. Hence, it can be concluded that the Random Forest Model is the most appropriate model for this dataset and should be used for predicting housing prices.

Combining Machine Learning Models to Predict House Prices