See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/383057414

House Price Prediction Using Machine Learning and Artificial Intelligence

Article in Journal of Artificial Intelligence & Cloud Computing · August 2024

DOI: 10.47363/JAICC/2024(3)357

CITATIONS READS

0 1,039

2 authors:

Fatbardha Maloku Besnik Maloku

Golden Gate University Golden Gate University
5 PUBLICATIONS 0 CITATIONS 4 PUBLICATIONS 0 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Fatbardha Maloku on 12 August 2024.

The user has requested enhancement of the downloaded file.

 ISSN: 2754-6659

Journal of Artificial Intelligence &

Cloud Computing

Review Article Open Access

House Price Prediction Using Machine Learning and Artificial

Intelligence
Fatbardha Maloku*, Besnik Maloku and Akansha Agarwal Dinesh Kumar

Master of Science in Business Analytics Student Candidates, Ageno School of Business, Golden Gate University, San Francisco, California 94105, USA

ABSTRACT
The escalating annual rise in housing prices introduces volatility and uncertainty into the real estate market, underscoring the critical need for accurate
price forecasting systems. Predicting house prices accurately remains challenging due to the multitude of influencing factors. This study aims to identify
and analyze key determinants affecting house prices, employing two established machine learning models. Through comparative analysis, the research will
recommend the most effective model for enhancing the accuracy of house price predictions.

*Corresponding author
Fatbardha Maloku, Master of Science in Business Analytics Student Candidates, Ageno School of Business, Golden Gate University, San Francisco, California
94105, USA.

Received: July 04, 2024; Accepted: July 08, 2024; Published: August 12, 2024

Introduction Problem Statement

The majority of people today are engaged in the commercial Business Problem Statement
activity of investing. Stocks, bonds, retirement, education, As the housing market is prone to volatility and uncertainty, it's
and other choices are widely used as investments. One of the critical to figure out which key metrics influence house price
investment forms that people frequently use is the buying of a predictability. House prices are commonly assumed to be tied to
property. The process is not simple, despite appearances to the our economy, but is this true? Despite the vast amount of data
contrary. Any real estate project that is purchased or in which an available, reliable property price projections are lacking.
investment is made sometimes necessitates a series of discrete
transactions involving numerous parties. As a result, it might be a Business Problem Background
crucial decision for both households and businesses. The housing In this study, we will examine and comprehend how various
market is currently being impacted by high-interest rates, which features might forecast house prices using the House Price
have raised home prices and affected both the supply and demand Prediction dataset. How do diverse characteristics like location,
for homes. Because of this, it is crucial to examine additional the house size, age, etc. affect the price of houses? The housing market
key metrics or factors that affect home prices. The purpose of this is currently being affected by the high-interest rates, which have
study is to forecast home values using two well-known machine boosted home prices and affected both the supply and demand for
learning models. Using the House Price Prediction dataset, we will homes. Analysis of other non-economic factors that affect housing
investigate and comprehend how different variables may forecast prices is crucial for this reason. When analyzing the price of a
home values. We will learn the impact of different factors like property, this analysis will assist buyers and sellers in focusing
location, size, house quality, condition etc. on the cost of homes. on some non-economic factors.
One of the various techniques for determining the value of a home
is prediction analysis. We will utilize both linear regression and Project Aim
random forest regression in this study to forecast house prices that This project's primary goal is to develop a Python-based machine
take other aspects into account. The knowledge obtained from this learning model that can learn from this data and calculate the
research will help customers decide when is the best time to buy cost of a home in any district given all of the other variables in
a home as well as real estate investors. the dataset.

Methods & Methodologies

The methodologies used in this project will start by running an
explanatory data analysis (EDA). We will continue to prepare the
dataset characteristics for use in model predictions and test a few
models to predict home prices in this project.

Exploratory Data Analysis (EDA)

The exploratory data analysis methodology helps us develop task
understanding, which in turn provides us with insights on later

J Arti Inte & Cloud Comp, 2024 Volume 3(4): 1- 10

Citation: Fatbardha Maloku, Besnik Maloku Akansha Agarwal Dinesh Kumar (2024) House Price Prediction Using Machine Learning and Artificial Intelligence.
Journal of Artificial Intelligence & Cloud Computing. SRC/JAICC-374. DOI: doi.org/10.47363/JAICC/2024(3)357

feature engineering and handling of missing values. Random Forest Regression

Random Forest Regression is a supervised learning algorithm.
Data Extraction This method combines predictions from multiple machine learning
This dataset was made publicly available on Kaggle, a website algorithms to make a more accurate prediction. A Random Forest
for data science competitions. To begin the analysis, we extracted is run by constructing several decision trees at the training time and
the data and put it into our notebook. outputting the mean of the classes as the prediction of all the trees.
The random Forest Regression model is powerful and accurate.
Data Formatting It is a great solution to many problems, including features with
The nature of the data and the relationship between the variables non-linear relationships [2].
should be examined and investigated before creating the prediction
models for estimating house prices. After importing the data The steps involved are described thoroughly in the below section.
into the environment, the attributes are then examined, including
names, data types, and the number of missing values. Identify your Dependent (y) and Independent Variables
(X)- Split the Dataset into the Training Set and Test Set-The
Data Preparation importance of the training and test split is that the training set
To assess the effectiveness of the final model, the data are divided contains known output from which the model learns. The model's
into a train set and a test set. Additionally, the data should be predictions are then put to the test using the knowledge it gained
formatted to make it simple to read and edit. This process should from the training set.
be done for each column and row in the dataset.
Training the Random Forest Regression Model on the Whole
Data Cleaning Dataset - The parameter n estimators create the n number of
After we prepare the data, we then need to decide how we will trees in your random forest, where n is the number, you pass in.
deal with the missing data. There are many variables with missing We passed 10. With the help of the fit () function, we can train
values. We attempt to impute the missing data using the following the model and improve accuracy by changing the weights in
methods: accordance with the data values. The predict () method is used to
• When dealing with numerical data, depending on the make predictions once our model has finished training.
distribution of the variable, we replace the missing values
with the mean or median. Predicting the Test Set Results-After successfully creating the
• For categorical data, we replace the missing values with 'N/A'. Random Forest Regression model, we can assess the accuracy
by calculating the R².R² score tells how the model is fitted to the
Visualize Numerical Variables data by comparing it to the average line of the dependent variable.
We may now begin visualizing our data. To examine the When the score is closer to 1, it shows the model performs well,
distributions, we first plot the histograms for each numerical when it is farther away from 1, it indicates that the model is not
variable. We are interested in the distributions' skewness or performing well.
symmetry as well as any other effects.
Solution Process
Correlation between the Explanatory Variables & Target By conducting a descriptive analysis of the data, we will begin
The relationship between these numerical variables and the solving the house prediction problem. We will learn information
target will next be examined. As a result, to see the associations from the descriptive analysis of the house prediction data set that
between the variables, we generate a correlation matrix and plot is not apparent from a simple glance at the spreadsheet. More
scatterplots. information about the house forecast's metadata is provided in
the section below.
Machine Learning Models
Linear Regression Descriptive Analysis
Linear Regression is a machine learning algorithm based on The house prediction data set's primary columns are summarized
supervised learning. It performs a regression task. Regression as follows:
models a target prediction value based on independent variables. • MSSubClass: Describes the kind of residence that is being
It is mostly used for finding out the relationship between variables sold.
and forecasting [1]. • LotFrontage: Linear feet of street connected to the property.
• LotArea: Square footage of the lot.
A linear model assumes a linear relationship between the input • Utilities: Available types of utilities.
variables (x) and the single output variable (y). More specifically, • Neighborhood: Physical locations within the city limits.
that y can be calculated from a linear combination of the input • BldgType: Residence type.
variables (x) [1]. • OverallCond: Evaluates the home's general condition.
• YearBuilt: Date of initial construction.
While training the model we are given: • YearRemodAdd: Remodel date.
x: input training data (univariate – one input variable(parameter)) • ExterCond: Assesses the outer material's state at the moment.
y: labels to data (supervised learning) • BsmtCond: Considers the basement's overall state.
• TotalBsmtSF: The sum of the basement's square feet.
The best line to predict the value of y for a given value of x is fitted • Heating: Heating type.
to the model during training. By determining the ideal values for • Fireplaces: Number of fireplaces.
the 1 (intercept) and 2 (coefficient of x), the model produces the • Bed: Number of bedrooms
best regression fit line. Once we find the best θ1 and θ2 values, we • Bath: Number of bathrooms
get the best fit line. So, when we are finally using our prediction • YrSold: Year Sold (YYYY)
model, it will predict the value of y for the input value of x. • SaleType: Type of sale
J Arti Inte & Cloud Comp, 2024 Volume 3(4): 2-10
Citation: Fatbardha Maloku, Besnik Maloku Akansha Agarwal Dinesh Kumar (2024) House Price Prediction Using Machine Learning and Artificial Intelligence.
Journal of Artificial Intelligence & Cloud Computing. SRC/JAICC-374. DOI: doi.org/10.47363/JAICC/2024(3)357

• SalePrice: The sale price of the houses min, and a max of our data set as well as the summary statistical
analysis of the columns in the house price data collection. By
The data set has a total of 30 columns and 1460 rows. The vast doing explanatory data analysis we discover that the maximum
majority of the data set's columns are classified as explanatory sales price for a house is $755,000, the minimum is $349,000
variables. The column labeled "Sales Price" will serve as and the average price of a house is approximately $163,000. We
the analysis' predictive variable. We will examine how these can see that there is a total of 30 features or variables, 8 of which
explanatory variables or factors affect the prediction of a house's are 'objects', 20 of which are 'int64', and 2 which are ‘float64'.
sales price during this analysis. To know more about the statistical According to the documentation of the competition, the variable
values of our dataset, we can use Python functions and methods to 'SalePrice', which has a data type of 'int64', is the target or label
gain more insight. In the example below, we've used the describe that we are going to predict.
function to learn more about the count, mean, standard deviation,

We learned during analysis that the dataset's neighborhood is a key variable. We can observe from the visualization below that the cost
of houses varies depending on the neighborhood within the same city. From the visualization, we can notice the diverse distribution
of house values among neighborhoods.

What time of year is ideal for selling a home? According to the analysis from real estate research company ATTOM Data Solutions,
late spring and early summer are the greatest seasons of the year to sell a home [3]. However, our house prediction dataset goes a
bit more in- depth and shows us the home sales for a period of four years, from 2006 through 2020. From the graph, we can see an
increase in the Sales Price in 2007 and a huge decrease in the market in the upcoming year, 2008. We understand that year 2007 had
the highest sales prices in the market, whereas in the upcoming year the prices shrank to approximately $8,000.

J Arti Inte & Cloud Comp, 2024 Volume 3(4): 3-10

Citation: Fatbardha Maloku, Besnik Maloku Akansha Agarwal Dinesh Kumar (2024) House Price Prediction Using Machine Learning and Artificial Intelligence.
Journal of Artificial Intelligence & Cloud Computing. SRC/JAICC-374. DOI: doi.org/10.47363/JAICC/2024(3)357

The descriptive analysis comes alive when there are distribution graphics of the explanatory variables in the relationship with the
Sales Price predictive variable. In this way, we will get to visualize how our data is distributed before predicting any values. To do
that, we have created a function which will go through explanatory columns in the house price dataset and visualize the results.

The next set of visuals demonstrates the distribution of MSSubClass, Utilities, and BldgType variables. From the chart, we can see
the distribution of our data in different categories. The MSSubClass that has the highest values is the class of 60. The utilities that
are highly used are the houses that include all public utilities such as (E, G, W, & S). The building type is approximately distributed
the same between single-family houses and townhouse end units.

When it comes to the general state of the houses, when the condition of the houses is 9 or above, the sales price of the houses tends
to be higher. The exterior condition of the property is crucial, and when it receives top ratings and presents a better impression, the
sales price of the home tends to be greater. Conversely, when the exterior condition of the home is poor, the sales price of the home
is lower. The general condition of the basement is another factor which plays a high role in the overall sales price of the house.

J Arti Inte & Cloud Comp, 2024 Volume 3(4): 4-10

Citation: Fatbardha Maloku, Besnik Maloku Akansha Agarwal Dinesh Kumar (2024) House Price Prediction Using Machine Learning and Artificial Intelligence.
Journal of Artificial Intelligence & Cloud Computing. SRC/JAICC-374. DOI: doi.org/10.47363/JAICC/2024(3)357

When the condition of the basement is good then the price of the house tends to be higher than in the houses where the condition of
the basement is poor.

The heating factor is also important. The gas-forced warm air furnace heating option is highly expensive compared to the floor furnace
and gravity furnace. The houses tend to be higher in price when the houses contain a central air conditioning system installed in place,
versus the ones that don’t. Houses that tend to have a higher number of fireplaces are also more expensive.

The garage and the number of parking spaces are two other factors that consumers consider when purchasing a home. We learned
through the analysis that homes with three garages are the most popular among buyers. Additionally, more expensive are homes with
555 square feet of pool space. Don't overlook the importance of the month that the residences were sold.

As we can see, the distribution is quite consistent over the entire month. We noticed a slight difference in the rate of price increases
in September.

The box plot visualization type can be used to display the distribution of the sales type over the sales price variable. We can observe
that the distribution changes between various categories from the graphic below. We can see the category “New” is a category for
the houses that are just constructed. This category seems to be higher distributed than the other categories of the types.

J Arti Inte & Cloud Comp, 2024 Volume 3(4): 5-10

Citation: Fatbardha Maloku, Besnik Maloku Akansha Agarwal Dinesh Kumar (2024) House Price Prediction Using Machine Learning and Artificial Intelligence.
Journal of Artificial Intelligence & Cloud Computing. SRC/JAICC-374. DOI: doi.org/10.47363/JAICC/2024(3)357

The density distribution of the Sales Price variable is shown in the below graph. The Sales Price target variable has a right-skewed
distribution and is not symmetric, which would have an impact on the model, according to the plot below.

Predictive Analysis
Predictive analytics, a subset of artificial intelligence, is a statistics-based technique that data analysts use to formulate hypotheses
and analyze past data in order to estimate the chance of a specific future result. Using machine learning and historical data like trends
and behaviors, predictive analytics enhances processes. On a set of future data, predictions are made using both predictive analytics
and machine learning. In our research, we used machine learning models to predict the possible house price based on the different
influential factors which we discussed in the above section.

Before building our machine learning models, we started by running a quick correlation test between variables to identify the highest
influencing variables which have a significant effect on house prices. As mentioned earlier, we have considered all the non-economic
factors to see which of these influences house prices the most.

J Arti Inte & Cloud Comp, 2024 Volume 3(4): 6-10

Citation: Fatbardha Maloku, Besnik Maloku Akansha Agarwal Dinesh Kumar (2024) House Price Prediction Using Machine Learning and Artificial Intelligence.
Journal of Artificial Intelligence & Cloud Computing. SRC/JAICC-374. DOI: doi.org/10.47363/JAICC/2024(3)357

Correlation Analysis
Correlation analysis is a test to examine the relationship between the variables in a dataset. With the correlation heatmap and correlation
matrix, one can observe the relationship between the explanatory and the response variables. For our study, we used a correlation
heatmap to examine the relationship between the variables.

As we can observe from the above heatmap most of the variables have a positive correlation with the sales price which is also known as
the house price in our study. Variables like lot area, house condition, basement condition, garage area etc. have a significant relationship
with the house prices. We will be taking all of these variables and proceeding to build our machine learning predictive models.

Linear Regression Model

We now move to the most important part of our analysis which is building the machine learning models to predict future outcomes.
We used our dataset to train the machine learning models to predict house prices. We started by building the Linear Regression Model
and trained it to help predict future outcomes [4].

We first used the algorithm to derive our x and y. We removed the garage year built and lot frontage, as these are already covered as
a part of the variable house-built year and the square feet of the entire house. We derived the sales price as y and other independent
variables as x. We then split the data into train and the test set as 0.80 and 0.20 as shown below.

J Arti Inte & Cloud Comp, 2024 Volume 3(4): 7-10

Citation: Fatbardha Maloku, Besnik Maloku Akansha Agarwal Dinesh Kumar (2024) House Price Prediction Using Machine Learning and Artificial Intelligence.
Journal of Artificial Intelligence & Cloud Computing. SRC/JAICC-374. DOI: doi.org/10.47363/JAICC/2024(3)357

After deriving the train and test set, we then trained our model and tested the training and testing accuracy. According to the results,
the training set has an accuracy of 79.44% as shown below, whereas the testing set accuracy score is 58.89% which is not bad.

Random Forest Regressor

To test and see if we can get a better accuracy score with other machine learning models, we decided to use the Random Forest
Regressor Model. A random forest is a meta estimator that employs averaging to increase predicted accuracy and reduce overfitting
after fitting multiple categorizing decision trees to different subsamples of the dataset.

We used the same split set to build a random forest regressor model and train the model. According to this model results, the training
set has an accuracy score of 97.32% and the test set has an accuracy of 82.29% as shown below.

J Arti Inte & Cloud Comp, 2024 Volume 3(4): 8-10

Citation: Fatbardha Maloku, Besnik Maloku Akansha Agarwal Dinesh Kumar (2024) House Price Prediction Using Machine Learning and Artificial Intelligence.
Journal of Artificial Intelligence & Cloud Computing. SRC/JAICC-374. DOI: doi.org/10.47363/JAICC/2024(3)357

As we can observe, clearly the random forest regressor model regression model with only 79.4% and 58.9% for training and
is better in comparison with the linear regression model. The testing, respectively. Therefore, we can conclude that the Random
reason for achieving a higher accuracy score is to ensure that Regression model is the best model for our study which can be
the predictive outcome is more accurate. Therefore, the random used to predict house prices with higher accuracy, given considers
forest regressor model is the clear winner in this case. Further, the same variables.
this trained random forest regressor model can be used to test the
dataset which consists of similar variables. Conclusion
For many years, house prices have been interpreted through
Model Results numerical values that contain various information about the
The analysis provided us with a collection of new information. houses. The use of statistical data causes an increase in the
Initially, the data contained both categorical and continuous number of scientific research based on housing data. One of the
features, and the target feature had a binary value. The data types frequently studied topics in these scientific studies is the prediction
for feature values are a mix of int, float, and object. Numerous of house prices. Our study is an example of these studies. The
columns had a significant number of missing values. Most house price prediction model uses machine learning algorithms
continuous feature variables have outliers, which we dealt with and models. We built two popular machine learning models to
during data pre-processing. Based on the heatmap and plot graphs, train the dataset. We measured their accuracy score on both the
there are dependent features that are closely associated with other training and testing set. Based on the accuracy score, we strongly
dependent features. In our analysis, we also noticed the diverse recommend a random forest regression model for better prediction
distribution of house values among neighborhoods. When it comes of house prices.
to the general state of the houses, when the condition of the houses
is 9 or above, the sales price of the houses tends to be higher. The model will be able to predict house prices exactly so that
buyers or sellers will not get lost. This will be useful to both
The exterior condition of the property is crucial. We learned buyers and sellers as well as real estate agents and companies.
through the analysis that homes with three garages are the most
popular among buyers. We noticed a slight difference in the rate Recommendations
of price increases in September. The sales price target variable As mentioned above, the Random Forest Regression model has
had a right-skewed distribution which affects the machine learning the highest accuracy score in terms of both training and test set. To
model. increase the prediction successes obtained in the study, enrichment
can be performed both in the study dataset and in the methods
The machine learning model used in this paper is the linear used. Many other variables affect house prices but are not taken
regression model and the random forest model. We considered into account. Variables such as location, room size, bathroom size
all the variables when training our model, except for garage year etc. that are ignored in most scientific studies can also directly
built and lot frontage as they are already a part of the house affect the results. For these reasons, instead of using a dataset
build year and the size of the house. We proceeded with all other consisting only of limited variables a large dataset can be used,
variables and trained our model, as we saw both the models had which also includes variables such as those mentioned here. In
a high accuracy score, however, the Random Forest Regressor addition, there are several other methods which can be used to
model had a higher accuracy score with 97.3% for the training increase the accuracy and prediction.
set and 82.3% for the testing set, in comparison with the linear

J Arti Inte & Cloud Comp, 2024 Volume 3(4): 9-10

Citation: Fatbardha Maloku, Besnik Maloku Akansha Agarwal Dinesh Kumar (2024) House Price Prediction Using Machine Learning and Artificial Intelligence.
Journal of Artificial Intelligence & Cloud Computing. SRC/JAICC-374. DOI: doi.org/10.47363/JAICC/2024(3)357

References 3. (2022) Home sales report. ATTOM https://www.attomdata.

1. (2022) Linear Regression. GeeksforGeeks ML https://www. com/solutions/market-trends-data/home-sales-report.
geeksforgeeks.org/ml-linear-regression/. 4. Brownlee J (2020) Linear Regression for Machine Learning.
2. Bakshi C (2022) Random Forest Regression - Level Up Machine Learning Mastery https://machinelearningmastery.
Coding. Medium https://levelup.gitconnected.com/random- com/linear-regression-for-machine-learning/.
forest-regression-209c0f354c84.

Copyright: ©2024 Fatbardha Maloku. This is an open-access article

distributed under the terms of the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium,
provided the original author and source are credited.

J Arti Inte & Cloud Comp, 2024 Volume 3(4): 10-10

View publication stats