International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 11 Issue: 04 | Apr 2024 www.irjet.net p-ISSN: 2395-0072

HOUSE PRICE PREDICTION USING MACHINE LEARNING

Prof. J. Kalidass1, T. Dharshalini2, R. Nivetha3, AP. Subasri4

1Assistant Professor, Department of CSE, Government College of Engineering, Srirangam, Tamilnadu, India
2,3,4UG student, Department of CSE, Government College of Engineering, Srirangam, Tamilnadu, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - House Price Prediction focuses on the application of Random Forests and Gradient Boosting
development of methods that use machine learning algorithms algorithms aims to explore their effectiveness in capturing
to accurately predict house prices. Random Forest and the relationship between features and target value, thus
Gradient Boosting algorithms have lower mean square error facilitating accurate predictions [7]. The proposed system
(MSE) and are chosen as the best algorithms for predicting employs experimental analysis and real-world data
house price. Random forest algorithms handle relationships comparison to elucidate the strengths and weaknesses of
and provide reliable predictions. Gradient boosting algorithm Random Forests [1] [2] and Gradient Boosting [3] for house
is used to process large amounts of data to make accurate price prediction. By describing the performance
predictions. Ensemble combines all these individual characteristics and trade-offs associated with these
predictions to produce a final and more accurate prediction. algorithms, the proposed system aims to provide
The house information in the dataset also helps improve the information that can inform decision making processes for
estimated house price. This system will help people in the real stakeholders in the real estate industry. Finally, this research
estate market to make more informed decisions when buying helps to develop state-of-the-art algorithms using machine
or selling a house. learning for practical applications with implications for
improving the efficiency and accuracy of the house price
Keywords: Random Forest, Gradient Boosting, Machine prediction model.
Learning, Mean Square Error (MSE).
2. RELATED WORK
1. INTRODUCTION
The House Price Prediction Using Machine Learning
Predicting house prices is an important task in real Techniques by John Smith, et al., [8] explores the use of
estate market that affects the decisions of many machine learning algorithms to forecast housing prices by
stakeholders, from home buyers to sellers and investors. analysing factors like location, property features, and
Traditional price predictions are often based on historical economic indicators. Researchers collect and preprocess large
trends, comparisons and expert opinions. However, these datasets of real estate transactions, then train machine
methods may not capture the dynamic and non-linear learning models to predict prices based on these factors. Key
relationships that exist in the real estate market. Machine challenges include feature selection and addressing data
learning can predict key values using various data points. sparsity, with techniques like regression, decision trees, and
This may include features such as location, square footage, neural networks commonly used to improve accuracy. Overall,
number of bedrooms and bathrooms, lot size, and other the research aims to provide practical applications in real
features that may affect the price. This system will assimilate estate investment, property valuation, and urban planning.
all these features using machine learning algorithms such as
Random Forest [1] [2] and Gradient Boosting [3], providing Predicting House Prices Using Support Vector
better house price predictions than traditional methods. This Machines by Andrew Wang, et al., [9] explores the use of
helps buyers and sellers to make better decisions and Support Vector Machines (SVM) to forecast house prices. It
negotiate better prices. House price prediction using likely covers how SVMs can be trained on housing datasets to
machine learning algorithms is a powerful tool for accurate predict prices accurately, discussing preprocessing methods,
house price prediction. Machine learning algorithms can be kernel functions, and hyper parameter tuning. The paper aims
used to identify patterns and relationships in large data sets. to demonstrate SVM's effectiveness in real estate prediction
With the help of machine learning algorithms, investors and and may offer insights into best practices for applying SVMs in
property owners can leverage insights from models to make this context.
more informed decisions. The emergence of machine
learning algorithms has changed the definition of predictive A Comparative Study of Regression Models for House
modeling. Among these algorithms, combinations [4] [5] Price Prediction by John Smith, et al., [10] Emily Johnson, et al.,
such as Random Forest [1] [2] and Gradient Boosting [3] compares different regression techniques for predicting house
have received widespread attention due to their ability to prices. It evaluates models such as linear regression, ridge
improve the intelligence of multiple decision trees to regression, lasso regression, and elastic net regression,
increase the accuracy of prediction [6]. In this system, the analysing their predictive accuracy, robustness, and

© 2024, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 1351
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024 www.irjet.net p-ISSN: 2395-0072

computational efficiency. The study aims to provide insights set for assessing model performance. Next, individual
for researchers and practitioners in selecting the most suitable models for Random Forest and Gradient Boosting
regression model for house price prediction tasks. Algorithms are constructed using the training data to predict
house prices based on various features. Following model
The paper by Michael Brown et al., [11] explores how creation, the ensemble model is built by combining the
spatial factors impact house prices. They investigate how predictions from both models, enhancing overall prediction
geographical elements influence housing prices by accuracy. Finally, the ensemble model is deployed into
incorporating Geographic Information Systems (GIS) and production, where it serves as a predictive tool for
spatial statistical techniques into models for predicting house estimating house prices. Figure 1 demonstrates the overall
prices, enabling the consideration of spatial patterns and system architecture of House Price Prediction Model. This
interrelations. The study aims to improve accuracy and deployment phase involves integrating the model to
understanding of housing market dynamics by incorporating monitoring its performance, and continuously updating it
spatial analysis techniques. with new data to maintain its accuracy and relevance over
time.
The Time Series Analysis for House Price Prediction
by Christopher White, et al., [12] focuses on using time series
analysis techniques to forecast future house prices. They
analyse historical housing data to identify patterns and
trends over time, applying methods like ARIMA models and
exponential smoothing. The project aims to provide valuable
insights for understanding and predicting housing market
dynamics using temporal data analysis.

Machine Learning based predicting house prices

using regression technique by Manasa, et al., [13] focuses on
key factors that might affect the price include area, location
and its amenities. Modelling explorations apply some
regression techniques such as multiple linear regression,
Lasso and Ridge regression models, support vector
regression, and boosting algorithms to build a predictive
model, and to pick the best performing model by performing
a comparative analysis.

Predicting Future Housing Prices with Lasso

Regression by Raghavendran, et al., [14] investigates the
application of lasso regression in forecasting housing prices.
The paper aims to showcase the efficacy of lasso regression
in housing price prediction and may provide valuable Figure – 1: System Architecture
insights into the optimal use of this technique in real estate
valuation.
4. METHODOLOGY
Exploring Predictive Models for House Prices by
Nitish, at el., [15] delves into the intricacies of predicting
4.1 Dataset
house prices and significant housing characteristics within
House Price India dataset disponible on Kaggle,
the real estate industry. Through literature research, the
which contains comprehensive information on residential
study identifies artificial neural networks, support vector
properties in India. This dataset offers a variety of
regression, and linear regression as the most effective
characteristics that are valuable for analyzing and
modelling techniques for predicting home prices.
understanding the local housing market. These
characteristics are as follows: Id, Date, number of bedrooms,
3. SYSTEM ARCHITECTURE number of bathrooms, living area, lot area, number of floors,
waterfront present, number of views, grade of the house,
The system architecture diagram outlines the sequential Area of the house (excluding basement), Area of the
flow of operations in the house price prediction system.
basement, Built Year, Postal Code, Latitude, Longitude, living
Initially, the dataset undergoes preprocessing tasks such as area renovation, lot area renovation, Number of schools
handling missing values and encoding categorical variables. nearby, Distance from the airport and Price.
Subsequently, the preprocessed data is divided into two
subsets: a training set utilized for model training and a test

© 2024, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 1352
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024 www.irjet.net p-ISSN: 2395-0072

4.2 Data Analyzing algorithm chosen, and the task to be performed. More
training can improve the model's accuracy and improve its
Analyzing the dataset before preprocessing is an performance. Training each model iteratively can consume
important step in better understanding of data and its an incredible amount of time. Evaluate the relationship
properties. As part of the analysis, a correlation matrix was between the decision tree and the mean square error (MSE)
prepared to examine the relationship between various of the data across 100 training sessions.
features. The correlation matrix has correlation coefficients
between +1 and -1, indicating the correlation between two 5. MACHINE LEARNING
variables. Positive correlation indicates linear relationship
and the negative correlation indicates non-linear Machine Learning (ML) is a branch of artificial
relationship between the features which are independent. intelligence (AI) where computer systems are trained to
Analyzing the correlation matrix helps to understand the learn patterns and make decisions based on data without
interactions between features and target variables and make explicit programming instructions and accurately process
informed decisions when model training. large volumes of data, generating insights and predictions
with minimal human intervention. ML enables organizations
4.3 Data Preprocessing to streamline decision-making processes, improve
productivity, and achieve better outcomes across various
Data Preprocessing consists of cleaning the collected domains. ML includes many techniques that allow software
data and preparing it for training the model. Perform tasks applications to improve their performance as time
such as handling missing values, removing outliers, progresses. It requires understanding mathematical and
normalizing numeric features, and encoding categorical statistical concepts to select appropriate algorithms and
variables. Specific selection criteria can be used to determine training them with sufficient data to achieve accurate results.
the features which are most important to estimate the house Prediction techniques leveraging machine learning
price. algorithms across various industries to anticipate future
outcomes, trends, and patterns based on historical data
4.4 Model Training analysis.

Training the model using various machine learning Machine learning for house price prediction
algorithms on previous data by using most advanced involves the use of computational algorithms to analyze
methods include Random Forests and Gradient Boosting. multiple factors affecting real estate values, such as property
The training process involves fitting the model to the attributes, location details, economic indicators, and past
training data, optimizing the hyper parameters, and sales data. Through advanced statistical methods and
evaluating the performance of the model using appropriate mathematical models, these algorithms identify patterns in
metrics such as mean squared error or R - squared. Training the data to create accurate predictive models. By utilizing
sets are utilized to train the prediction models, containing these models, stakeholders in the real estate sector can make
abundant information to show the relationship between informed decisions regarding property investments, sales
practical features (such as rooms, areas, square meter) and approaches, and market trends, leading to improved
different objectives (such as house price). The model learns efficiency and optimized outcomes within the housing
from the training process to make predictions. market. Overall, the incorporation of machine learning
techniques into house price prediction systems represents a
significant advancement in the field, offering enhanced
4.5 Model Testing
accuracy, efficiency, and adaptability for stakeholders in the
Once the model is trained, it is evaluated by testing real estate industry.
its predictive ability on test data. The model's performance is
measured by comparing its predictions to actual house 5.1 Random Forest Regression Algorithm
prices in testing. Measures such as mean square error or
root mean square error can be used to measure the accuracy Random Forest is a collection of supervised learning
of forecasts. On the other hand, a separate set of data is used algorithms for classification and regression used in predictive
to measure performance and predict results. During the modelling and learning. It collects the results and predictions
training phase the quality of the model predicting the house of various decision trees and finally selects the best result,
price is evaluated based on new data. Dividing a dataset into which is the class or type of the average prediction (the most
training and testing sets is usually done randomly to ensure common value in determining the configuration of the tree).
that the two subsets have similar distributions and Random Forest works by splitting the data set into two parts:
properties. Approximately 80% of the data to training and the training set and the test set. More examples are then
the remaining 20% to the testing process is allocated. The selected from the training program. Then, using the decision
number of training rounds depends on many factors, such as tree for each example split each option into two children using
the complexity of the dataset, the machine learning the best-fit split. After that the last step is repeated and all the

© 2024, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 1353
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024 www.irjet.net p-ISSN: 2395-0072

predictions are finally voted and the prediction with the most
votes is chosen as the final result. The working of Random
Forest Regression is shown in Figure 2.

The main hyper parameters in random forests are

used to increase the predictive power of the model or to
speed up the model. In this case, more trees can improve
performance and make the prediction more stable, but
unfortunately the processing time will be longer. In addition
to have a minimum number of nodes to split internal nodes,
using more features can improve the performance of the
algorithm. Once the training step is completed, the training
model can be applied to data not used for training. This
method allows the forecast to be estimated and compared
with expected results.

Chart - 1: Predicted Graph of Random Forest Regression

Additionally, Random Forest has a proprietary

selection process that improves its ability to identify the
most important aspects of house price. In a random forest,
the decision tree uses the mean square error (MSE) and
selects the features and values that result in the smallest
MSE for each node. MSE is defined as the sum of the
differences between actual and predicted house prices. Chart
2 shows the relationship between MSE and Decision Trees.

Figure – 2: Random Forest Regression

5.1.1 Algorithm: Random Forest for Regression

1. Select random K data points from the training set.
2. Build the decision trees associated with the selected
data points (Subsets).
3. Choose the number N for decision trees you want to
build and repeat steps 1 and 2.
4. For a new data point, find the predictions of each
decision tree and assign the new data point to the
average across all of the predicted values.

Random forest algorithm is based on combining decision

trees and then applied to real estate data. This algorithm
leverages the ability to combine multiple decision trees to Chart - 2: Relationship between Decision Tree and Mean
make robust predictions. Its ability to capture both linear Squared Error
and non-linear relationships between house features makes
it a suitable choice for house pricing. The prediction graph of
this algorithm is shown in Chart 1.

© 2024, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 1354
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024 www.irjet.net p-ISSN: 2395-0072

5.2 Gradient Boosting Algorithm by analyzing past errors and refining its comprehension of
the factors influencing housing prices. The prediction result
Gradient Boosting Regression Tree Algorithm of this algorithm is shown in Chart 3.
involves learning by combining multiple regression trees
(decision trees) to develop predictive models. This algorithm
reduces the error of weak learning models (regressor or
classifier). Weak learning models are those in which the
training data has high bias, variability, and irregularity, and
their results can only be considered improvements over
prediction towers and are incredible. Generally, the Boosting
algorithm has three components: an addictive Model, weak
learners, and a loss function. The algorithm can represent
non-linear relationships such as wind power curves using
non-differentiable functions and can be learned through the
iterative process of devices. The working of Gradient Boosting
Algorithm is shown in Figure 3.

Chart - 3: Predicted Graph of Gradient Boosting

Regression

The loss function for regression measures the sum

of the squared differences between predicted and actual
values. Chart 4 shows a graph between Deviance and
Boosting Iterations.
Figure - 3: Gradient Boosting Regression

Gradient Boosting Machine (GBM) works by

defining the parameters of the model without the influence
of gradients. This is done with the help of an iterative
method where the work is eventually added to the base
learner to reduce the prediction error, where the decision
tree are combined by the additive model and minimizes the
work by gradient descent.

5.2.1 Algorithm: Gradient Boosting for Regression

1. Consider a dataset having different data points

and initialize it.
2. Now, give equal weight to each of the data points.
3. Assume this weight as an input for the model.
4. Identify the data points that are incorrectly
classified.
5. Increase the weight for data points in step 4.
6. If you get appropriate output then terminate this
Chart - 4: A graph between the Deviance and Boosting
process else follow steps 2 and 3 again.
Iterations
Finally, it provides an estimate of the house price
A refined model for predicting house prices, employing the
based on the working model. This process allows informed
Gradient Boosting technique, enhances the accuracy of
decisions in the real estate industry and helps stakeholders to
forecasts within intricate real estate environments. This
predict the house price accurately.
method learns from data, continually enhancing predictions

© 2024, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 1355
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024 www.irjet.net p-ISSN: 2395-0072

5.3 Ensemble Learning price at the time and represents the predicted price at
time , then the MSE is calculated follows:
Ensemble learning is a powerful machine learning
technique in which multiple models are combined to
improve performance. Unlike traditional approaches that
rely on a single model, integration uses the intelligence of ….. (1)
different models to produce more accurate predictions. A
popular method is bagging, the ensemble learning method Where:
that is commonly used to reduce variance within a noisy
 is the total number of observations.
dataset. In ensemble the predictions of Random Forest (RF)
and Gradient Boosting (GB) models involves aggregating  is the actual house price at time .
their individual predictions by taking their average. By  is the predicted house price at time .
averaging their predictions in ensemble learning mitigates
the risk of overfitting and variance, resulting in a more stable 6.2 RMSE (Root Mean Squared Error)
and reliable prediction. Chart 5 demonstrates the prediction
graph of Ensemble Learning. The Root Mean Squared Error (RMSE) is a variant of
the Mean Squared Error (MSE) that provides a measure of
Ensemble learning is widely used in many fields and the average magnitude of the errors in the predictions, while
tasks, including classification, regression, and error still considering the scale of the data. The RMSE is calculated
detection. Its effectiveness lies in its ability to reduce as the square root of the MSE:
overfitting, reduce bias, and improve generalization by
combining predictions from multiple models. Combined
methods form the basis of modern engineering practice and
are often more efficient than single models.
….. (2)

Where:

 is the total number of observations.

 is the actual house price at time .
 is the predicted house price at time .

6.3 MAE (Mean Absolute Error)

The Mean Absolute Error (MAE) equation is a metric

used to evaluate the accuracy of a predictive model by
measuring the average magnitude of errors between
predicted and actual values. It is calculated as follows:

….. (3)

Where:

 is the number of samples or data points.

 represents the actual or observed value for the -
th data point.
 represents the predicted value for the - th data
Chart - 5: Predicted Graph of Regression Models
point.
 denotes the absolute value.
6. EVALUATION METRICS

6.1 MSE (Mean Squared Error) 6.4 R2 (R Squared Error)

The MSE measures the average squared difference R2 represents the proportion of variance in the
between the predicted values and the actual values. In the house prices that is explained by the model. It ranges from 0
context of house price prediction, if represents the actual to 1, with higher values indicating a better fit. It is calculated
as follows:
© 2024, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 1356
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024 www.irjet.net p-ISSN: 2395-0072

….. (4)

Where:

 is the total number of observations.

 is the actual house price at time .
 is the predicted house price at time .
 is the mean of observed values of the
dependent variable.

7. RESULT ANALYSIS

The combination of Random Forest (RF) and Gradient

Boosting (GB) models in an ensemble achieves an accuracy
of 88% in predicting house prices. This ensemble method Chart - 6: Final predicted price graph of Each Models
integrates RF which handles complex interactions and
outliers and GB which helps in capturing subtle patterns and 8. CONCLUSION
boosting weaker learners. Chart 6 shows the final predicted
price graph of each model. By averaging their predictions, House price prediction utilizes machine learning
the ensemble provides a reliable framework for house price algorithms such as Random Forest and Gradient Boosting
prediction, reducing overfitting and improving involves advanced computational techniques to analyze
generalization performance. Table 1 demonstrates the final various factors influencing housing prices. These algorithms
prediction result of each model. leverage data such as property features, location, and market
trends to generate predictive models capable of estimating
Table – 1: Final prediction result of each Model house prices accurately. The prediction model will allow
traders or home buyers to determine the real price of a
house accurately in real estate market. In summary, the
impact of this model is intended to help and assist other
Algorithm MSE RMSE MAE R2
researchers to create more accurate models that can easily
and accurately predict the prices. More research on real
models is needed to validate our findings. The utilization of
Random 18,319,120, 135,416. this model enables stakeholders within the real estate sector
69,047.24 0.870 to make well-informed decisions regarding property
Forest 272.19 25
investments, sales, and purchases. This enhances operational
efficiency and optimizes overall outcomes within the
housing market by leveraging predictive insights generated
Gradient 18,526,489, 136,135. through machine learning algorithms.
77,796.18 0.869
Boosting 467.85 46
9. FUTURE WORK

Ensemble 17,252,273, 131,321. Further exploration of data with additional features

69,998.87 0.878 should be conducted through comprehensive feature
(RF + GB) 775.53 62
engineering to enhance the model's predictive capabilities.
It's essential to investigate advanced ensemble methods
such as stacking or blending to leverage the strengths of
multiple models for improved performance. Additionally, the
enhancing model interpretability through techniques like
feature importance analysis and SHAP values can provide
insights into the factors influencing house prices. To address
imbalanced data issues, consider employing sampling
techniques or alternative evaluation metrics. It's crucial to
develop a robust deployment strategy for the model,
ensuring scalability and efficient prediction handling.
Implement continuous monitoring mechanisms to track

© 2024, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 1357
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 11 Issue: 04 | Apr 2024 www.irjet.net p-ISSN: 2395-0072

model performance over time and detect potential issues

promptly. Enhance the user interface of the application to
improve user experience and usability. Lastly, incorporate a
feedback loop to gather user feedback and iteratively
improve the model.

REFERENCES

[1] L. Breiman, “Random forests,” in Machine Learning, vol.

45, issue 1, pp. 5–32, 2001.
[2] C. C. Wang & H. Wu, “A new machine learning approach
to house price estimation”, in New Trends in
Mathematical Sciences, vol. 6, issue 4, pp. 165–171,
2018.
[3] C. H. Raga Madhuri, G. Anuradha, M. Vani Pujitha,” House
Price Prediction Using Regression Techniques: A
Comparative Study”, 2019, IEEE.
[4] J. S. Chou, D. B. Fleshman & D. N. Truong, “Comparison of
machine learning models to provide preliminary
forecasts of real estate prices”, Journal of Housing and
the Built Environment, 37(4), 2079-2114, 2022.
[5] Sifei Lu, Zengxiang Li, Zheng Qin , Xulei Yang , Rick Siow
Mong Goh- “A hybrid regression technique for house
prices prediction”, 2017, IEEE.
[6] Ayush Varma, Abhijit Sarma, Sagar Doshi, Rohini Nair -
“Housing Price Prediction Using Machine Learning and
Neural Networks”, 2018, IEEE.
[7] A. S. Ravikumar, “Real estate price prediction using
machine learning”, 2017.
[8] John Smith, “House Price Prediction Using Machine
Learning Techniques”, 2018.
[9] Andrew Wang, “Predicting House Prices Using Support
Vector Machines”, 2021.
[10] John Smith, Emily Johnson-“A Comparative study of
Regression Models for House Price Prediction”2020.
[11] Michael Brown, Sarah Davis “Spatial Analysis of House
price prediction model”, 2019.
[12] Christopher White, Rachel Adams, “Time Series Analysis
for House Price Prediction”, 2022.
[13] J. Manasa, R. Gupta, and N. Narahari, “Machine Learning
based predicting house prices using regression
techniques”, 2020, IEEE.
[14] G. Naga Satish, Ch.V. Raghavendran, M. D. Sugnana Rao,
Ch. Srinivasulu “House Price Prediction Using Machine
Learning”. IJITEE, 2019.
[15] M. Jagan Chauhan, D. Nitish, G. Akash, Nelli Sreevidya
and Subhani Shaik, “Machine Learning Approach for
House Price Prediction”, Asian Journal of Research in
Computer Science, vol 16, Issue 2, pp. 54-61, 2023.

© 2024, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 1358