INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 1714
www.rsisinternational.org
Predictive Modelling and Statistical Analysis of Housing Prices in
Lagos State, Nigeria
Odukoya E.A., Oyelakin O.P., Lawal O. J
Department of Statistics, Faculty of Science, Ekiti State University, Ado-Ekiti, Ekiti State, Nigeria
DOI: https://doi.org/10.51244/IJRSI.2025.120800153
Received: 12 July 2025; Accepted: 18 July 2025; Published: 16 September 2025
ABSTACT
This study delves into housing market analysis and price prediction, leveraging statistical modeling and
machine learning techniques to uncover patterns and forecast property prices. The housing market is
influenced by diverse factors such as location, property features, economic indicators, and market trends,
necessitating a comprehensive analytical approach. Using a dataset comprising historical housing prices and
relevant attributes, the study employs exploratory data analysis to identify key determinants of property values.
The findings highlight significant predictors of housing prices and demonstrate the potential of predictive
analytics in guiding buyers, sellers, and policymakers. This research offers valuable insights into market
dynamics and contributes to data-driven decision-making in real estate
Keywords: Machine Learning Techniques, Housing Market, Price Prediction
INTRODUCTION
The housing market which is important to the economy affects individual wealth and overall economic health.
It's almost vital for a range of people like potential buyers, investors, policymakers, and real estate experts to
get a good grasp on housing market patterns and predict property prices accurately. So basically, looking at the
housing market and trying to figure out future prices means you need to get what's going on in residential
properties.
Different analytical methods are employed to study and predict market behaviour. These methods range from
basic statistics, which provide simple metrics like average prices, to more sophisticated techniques such as
time series analysis and comparative market analysis. Advanced approaches also include machine learning
models that use large datasets to forecast future prices based on various factors. However, predicting housing
prices is challenging due to the volatility of the market and the complexity of factors involved. Changes in the
economy, government policies, and demographic shifts can all influence housing prices in ways that are not
always predictable. Overall, housing market analysis and price prediction are vital for making informed
decisions in real estate investment, policymaking, and valuation. They help stakeholders understand market
trends, anticipate future conditions, and navigate the complexities of the housing market effectively.
The housing market plays a crucial role in a country's economy and significantly impacts the financial well-
being of individuals and families. Housing prices are influenced by a multitude of factors, including local
market demand, the economic environment, interest rates, and government policies (Paciorek, 2013). As such,
understanding these influences and being able to predict housing prices is of great interest to economists,
policymakers, investors, and home buyers. The price of housing is not only a reflection of property
characteristics such as size, location, and amenities but also macroeconomic indicators like inflation,
unemployment rates, and mortgage interest rates (Kain & Quigley, 1970). Numerous studies have shown that
housing prices can be volatile, responding to sudden changes in these factors (Glaeser, Gyourko, & Saks,
2005). This volatility poses risks to stakeholders in the housing market, making accurate predictions of
housing prices critical for informed decision-making.
With advancements in data science and machine learning, predictive models for housing prices have improved
significantly. Machine learning algorithms can leverage vast amounts of historical data, property
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 1715
www.rsisinternational.org
characteristics, and economic variables to generate more accurate price forecasts (Bajari, Chu, & Park, 2012).
Traditional models, such as hedonic pricing models, use regression analysis to estimate the influence of
various factors on housing prices. However, machine learning techniques, like random forests, gradient
boosting, and neural networks, have been shown to outperform traditional models due to their ability to capture
nonlinear relationships between features (Galster & Tatian, 2009). Studies by Bajari et al. (2012) and Wen,
Yang, and Song (2019) highlight the effectiveness of machine learning models in predicting real estate prices.
These models consider an array of variables, including structural features (e.g., square footage, age of
property), location features (e.g., proximity to schools, transportation), and economic indicators (e.g., interest
rates, inflation). The integration of such variables into predictive models offers greater flexibility in adjusting
for changing market conditions and helps stakeholders make better-informed decisions. Even with these
improvements, there are still hurdles in creating dependable forecasting systems. Real estate markets tend to be
diverse, so the things that impact home prices in one area can be quite different from those in another. In
addition, the availability and quality of data might vary a lot, which could affect how well the forecasting
systems work, the housing data was analysed from townhouses in Fairfax Country and compared the
classification accuracy performance of various algorithms. To help a real estate agent, he then developed a
better prediction model for enhanced decisions based on house price assessment. Jafari and Akhavian (2019)
stated that the square footage of a unit of a house is the most important variable in predicting the price of a
house, followed by the number of bathrooms and number of bedrooms. Raga Madhuri, Anuradha, & Vani
Pujitha, (2019) discussed diverse regression techniques such as Gradient boosting and AdaBoost Regression,
Ridge, Elastic Net, Multiple linear, and Lasso to locate the most excellent. The performance measures used are
Mean Square Error (MSE) and Root Mean Square Error (RMSE). Predicting the failure of industrial
equipment’s mechanical parts Regression is a supervised learning technique that aims to find the relationships
between the dependent and independent variables. Ridge regression is a method of estimating the coefficients
of multiple regression models in scenarios where the independent variables are highly correlated (Hilt, &
Seegrist, 1977). It has been used in many fields including econometrics, chemistry, and engineering (Gruber,
& Schucany, 2020). Uzoma, & Jeremiah, (2016) developed outlier detection and optimal variable selection
techniques in regression analysis and other fascinating papers by the research include (Anabike et al., 2023;
Innocent et al., 2023; Abuh, Onyeagu, & Obulezi, 2023a; Abuh, Onyeagu, & Obulezi, 2023b; Obulezi et al.,
2022; Onyekwere, & Obulezi, 2022; Onyekwere et al., 2022). This section summarizes the concept of relevant
work on Prediction of House Prices in Lagos Nigeria using a machine learning model. Here, the house price
prediction can be divided into two categories (Zulkifley et al., 2020), first by focusing on house characteristics,
and secondly by focusing on the model used in house price prediction. Many researchers have produced a
house price prediction model, including Temur, Akgün, & Temur, (2019), Jafari, & Akhavian, (2019), Gao et
al.
METHOD
Data Collection Method
This study use secondary data obtained from several real estate’s company platforms specialized in the
purchasing and selling of land, building of houses (bungalow, apartments, shops, etc) in Lagos State, Nigeria.
Real Estate Platforms:
Extract data from popular property listing websites in Lagos (e.g., PropertyPro.ng, PrivateProperty.com.ng) to
obtain information on housing prices, features, and availability.
Government and Agency Reports:
Use reports and datasets from relevant Nigerian agencies like the Lagos State Ministry of Housing, National
Bureau of Statistics (NBS), and Central Bank of Nigeria (CBN) for economic and demographic data.
Historical Market Data:
Collect historical property transaction records, where available, from real estate firms or public registries
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 1716
www.rsisinternational.org
Data Analysis
The analysis of the data followed a multi-step process, utilizing both quantitative and qualitative methods to
interpret the information collected. The analytical approach used in this study included the following
techniques:
Simple Linear Regression Analysis
Linear regression assumes a linear relationship between the dependent variable 𝛾 (e.g., housing price) and
independent variables X (e.g., size, number of bedrooms, location). The goal is to find the best-fitting line (or
hyperplane in multiple dimensions) that predicts Y based on the X variables.
Model Representation
The relationship is typically expressed with the following equation:
𝜸 = 𝜷
𝟎
+𝜷
𝟏
𝑿+∈
Where
𝛾 : The dependent variable (the outcome being studied, such as hosing price).
𝛽
0
: The intercept (the value of 𝛾 when all independent variables are zero).
𝛽
1
: The coefficient of the independent variables. These represent the change in the dependent variable for a
one-unit change in the corresponding independent variable.
X: The independent variables (predictors), which may include property size, lot size or the dependent
variable.
∈: The error term, accounting for the variance in 𝛾 not explained by the independent variables. It captures
the factors that affect 𝛾 but are not included in the model
Model Form:
Price = 𝜷
𝟎
+ 𝜷
𝟏
(Size) + ϵ
Multiple linear regressions Analysis
Multiple Linear Regression (MLR) is a statistical technique used to model the relationship between one
dependent variable (in this case, housing price) and multiple independent variables (like square footage,
number of bedrooms, location factors, etc.). It is a common approach for predicting housing prices because it
can handle the complexity of multiple factors influencing a single outcome.
Model
𝜸 = 𝜷
𝟎
+𝜷
𝟏
𝑿
𝟏
+𝜷
𝟐
𝑿
𝟐
+…………….+𝜷
𝒏
𝑿
𝒏
+∈
Where
𝛾 : The dependent variable (the outcome being studied, such as maternal mortality ratio).
𝛽
0
: The intercept (the value of 𝛾 when all independent variables are zero).
𝛽
1
,𝛽
2
,…𝛽
𝑛
: The coefficients of the independent variables. These represent the change in the dependent
variable for a one-unit change in the corresponding independent variable.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 1717
www.rsisinternational.org
𝑋
1
,𝑋
2
,…𝑋
𝑛
: The independent variables (predictors), which may include socio-economic, healthcare access,
or demographic factors
∈: The error term, accounting for the variance in 𝛾 not explained by the independent variables. It captures
the factors that affect 𝛾 but are not included in the model.
Fit the Model: Use statistical software (e.g., Python’s scikit-learn or R) to fit the MLR model to the training
data, estimating coefficients for each independent variable.
Interpret Coefficients: Each coefficient 𝛽
1
indicates the effect of that variable Xi on Y, holding other
variables constant. For example, in a housing study, the coefficient for square footage shows how much price
is expected to increase per additional square foot.
Residual Analysis: Examine the residuals to ensure assumptions (like homoscedasticity and normality) are
met.
Model Form:
Price = 𝜷
𝟎
+ 𝜷
𝟏
(Size) + 𝜷
𝟐
(Bedrooms) + …+
Table 1 Descriptive Analysis
Sample
Size
Max
Min
Standard dev
Kurtosis
Skewness
Market
value
100
200000000
80000000
42163702.14
-0.971445666
0.544524852
Average
monthly
rent
100
1000000
400000
214617.348
-1.283177763
0.140134002
Property
age
100
14
2
4.680445028
-1.440113222
0.33644709
Power
supply
reliability
hrs per day
100
20
12
3.046358979
-1.423817641
-0.187987711
Distance
to closest
major road
km
100
2
0.5
0.56183322
-1.368045445
0
Number of
bedrooms
100
5
1
1.42133811
-1.304933726
0
Number of
bathrooms
100
5
1
1.42133811
-1.304933726
0
Parking
spaces
100
3
1
0.822597512
-1.522682477
0.018698722
Size sqm
100
300
100
71.06691
-1.30493
0
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 1718
www.rsisinternational.org
Interpretation
Market Value
The market value of properties ranges from 8,000,000 to 20,000,000, with an average value of 13,000,000. The
standard deviation of 42,163,702.14 indicates significant variability, reflecting a wide range of property values.
The positive skewness (0.5445) suggests that slightly more properties have market values below the average.
The kurtosis value (0.9714) is close to zero, indicating a near-normal distribution.
Average Monthly Rent
The monthly rent for properties varies between 400,000 and 1,000,000, with a mean rent of 680,000. The
standard deviation of 214,617.35 suggests moderate variability in rent prices. The skewness (0.1401) is close
to zero, indicating a nearly symmetric distribution, while the kurtosis (1.2832) shows a slightly peaked
distribution compared to a normal curve.
Property Age
The age of properties ranges from 2 to 14 years, with an average age of 7.25 years. A standard deviation of
4.68 years reflects moderate variability. The positive skewness (0.3364) indicates a slight tendency for
properties to be younger than the average. The kurtosis (1.4401) suggests a slightly more peaked distribution
than normal.
Power Supply Reliability (Hours per Day)
Power supply reliability ranges from 12 to 20 hours daily, with a mean of 16.25 hours. The standard deviation
of 3.05 hours indicates moderate variability. The skewness (-0.1880) is slightly negative, suggesting a small
tendency for properties to have power supply reliability above the average. The kurtosis (1.4238) reflects a
slightly peaked distribution.
Distance to Closest Major Road (km)
The distance to the nearest major road ranges from 0.5 km to 2 km, with an average of 1.25 km. A standard
deviation of 0.56 km shows low variability. The skewness is zero, indicating a perfectly symmetric
distribution, while the kurtosis (1.3680) suggests a slightly peaked distribution.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 1719
www.rsisinternational.org
Number of Bedrooms
Properties have between 1 and 5 bedrooms, with an average of 3 bedrooms. The standard deviation of 1.42
indicates moderate variability. Both skewness and kurtosis are zero, suggesting a symmetric and normal
distribution.
Number of Bathrooms
The number of bathrooms also ranges from 1 to 5, with an average of 3 bathrooms. The standard deviation
(1.42), skewness (0), and kurtosis (0) are identical to the number of bedrooms, indicating similar distributional
characteristics.
Parking Spaces
The number of parking spaces ranges from 1 to 3, with a mean of 1.99 spaces. The standard deviation of 0.82
reflects low variability. The skewness (0.0187) is close to zero, indicating an almost perfectly symmetric
distribution, while the kurtosis (1.5227) suggests a slightly peaked distribution.
Size (Square Meters)
Property sizes range from 100 to 300 square meters, with an average size of 200 square meters. The standard
deviation of 71.07 reflects moderate variability. The skewness is zero, indicating a symmetric distribution,
while the negative kurtosis (-1.3049) suggests a flatter (platykurtic) distribution compared to normal.
Table 2 Linear Regression
coefficients
Standard error
T stat
Intercept
38000000
8029283.141
4.732676546
Size sqm
460000
37850.40371
12.15310683
Interpretation
The regression results indicate a significant relationship between the size of an asset (measured in square
meters) and its estimated value. The model suggests that for every additional square meter of size, the value
increases by 460,000 currency units on average. This relationship is supported by a strong t-statistic of 12.15,
indicating high statistical significance.
The intercept of the model is 38,000,000, which represents the estimated value of the asset when its size is
zero. While this value may not have a practical interpretation in real-world terms, it serves as a baseline for the
regression equation.
Overall, the findings demonstrate that size is a crucial factor in determining the asset's value, and the model
provides a robust fit, as evidenced by the statistically significant coefficients for both the intercept and the size
variable.
Table 3 Multiple Regression
coefficients
Standard error
T stat
Intercept
-4031937.86
20372886.52
-0.19790705
Average monthly rent
174.7739625
4.934564662
35.41831437
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 1720
www.rsisinternational.org
Property age
1276.842391
1366661.045
0.000934279
Power supply reliability hrs per day
-5.97447e-10
392576.0752
-1.5219e-15
Parking space
31921.05978
882721.7619
0.036162085
Size sqm
75500.47462
14893.22067
5.069452491
The regression analysis provides insights into the factors influencing the market value of properties. Here is a
detailed interpretation of the results:
The intercept, with a coefficient of -40,319,387.86, represents the predicted market value when all independent
variables are zero. However, this value lacks practical meaning in this context, as it is unrealistic for properties
to have negative market values. The very low t-statistic (-0.1979) further suggests that the interception is not
statistically significant.
Among the independent variables, the average monthly rent stands out as a highly significant predictor of
market value. With a coefficient of 174.77, it indicates that for every 1-unit increase in rent, the market value is
expected to rise by 174.77 units. The small standard error (4.93) and the very high t-statistic (35.42) confirm
its strong predictive power.
Property age has a coefficient of 1,276.84, suggesting that for each additional year of age, the market value
increases by this amount. However, the large standard error (1,366,661.05) and extremely low t-statistic
(0.00093) indicate that property age is not a statistically significant factor in determining market value.
Power supply reliability, measured in hours per day, has a negligible coefficient (-5.97e-10), meaning it has no
meaningful impact on market value. The extremely low t-statistics (-1.52e-15) further supports its lack of
significance.
Parking spaces have a coefficient of 31,921.06, indicating a positive relationship with market value. However,
the high standard error (882,721.76) and low t-statistic (0.036) suggest that the effect of parking spaces is not
statistically significant in this model.
The size of the property, measured in square meters, is another significant predictor of market value. With a
coefficient of 75,500.47, it indicates that for every additional square meter, the market value increases by this
amount. The relatively low standard error (14,893.22) and high t-statistic (5.07) confirm its importance as a
predictor.
Correlation
Market
value
Average
monthly
rent
Propert
y age
Power
supply
reliability
hrs per day
Distance
to closest
major road
km+
Number of
bedrooms
Number of
bathrooms
Parking
spaces
Size sqm
Market
value
1
Average
monthly rent
0.98230
1
Property age
0
0
1
Power
supply
reliability
hrs per day
0
0
-0.199
1
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 1721
www.rsisinternational.org
Distance to
closest
major road
km
0
0
-0.984
0.33197
1
Number of
bedrooms
0.7753
0.7285
0
0
0
1
Number of
bathrooms
0
0
-0.984
0.33197
1
0
1
Parking
spaces
-0.0320
-0.0354
-0.012
-0.02721
0.00546
-0.00864
0.00546
1
Size sqm
0.7753
0.7285
0
0
0
1
0
-0.00864
1
Interpretation
The correlation matrix reveals several significant relationships between the property-related variables. A very
strong positive correlation exists between market value and average monthly rent (0.983), indicating that
properties with higher market values tend to command higher rents. Similarly, market value is strongly
associated with the number of bedrooms (0.775) and size in square meters (0.775), suggesting that larger
properties and those with more bedrooms are generally more valuable.
The relationship between the average monthly rent and the number of bedrooms is also strong (0.729),
showing that properties with more bedrooms typically have higher rental prices. Additionally, size in square
meters correlates strongly with the number of bedrooms (0.775), highlighting that larger properties tend to
accommodate more bedrooms.
Interestingly, there is a weak positive correlation between power supply reliability and distance to the closest
major road (0.332), which suggests that properties farther from major roads might have slightly more reliable
power supply. However, power supply reliability shows a weak negative correlation with property age (-
0.199), indicating that older properties may experience slightly less reliable power.
The number of bedrooms and number of bathrooms are perfectly correlated (1.000), implying that properties
with more bedrooms invariably have an equal number of bathrooms. On the other hand, the number of parking
spaces shows little to no correlation with other variables, indicating that it is an independent feature of the
property.
Overall, the strongest relationships involve market value, rent, size, and the number of bedrooms, while
variables like parking spaces and power supply reliability show weaker or negligible associations with most
other factors.
CONCLUSION
The study concludes that housing prices are primarily driven by economic and physical attributes, such as
rental income potential and property size. These factors reflect the utility and desirability of properties in the
market. On the other hand, aspects like property age, proximity to roads, and parking spaces may hold less
relevance in certain contexts or regions. The results underscore the importance of focusing on key variables
when analysing housing markets or predicting property prices.
REFERENCE
1. Abuh, J., Onyeagu, A., & Obulezi, M. (2023a). Regression modelling in predictive analytics.
2. Abuh, J., Onyeagu, A., & Obulezi, M. (2023b). Machine learning applications in econometrics.
3. Adams, Z., & Fuss, R. (2010). Housing prices and macroeconomic fundamentals: An empirical
analysis across countries. Journal of Housing Economics, 19(2), 84-103.
INTERNATIONAL JOURNAL OF RESEARCH AND SCIENTIFIC INNOVATION (IJRSI)
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue VIII August 2025
Page 1722
www.rsisinternational.org
4. Anabike, E., Innocent, I., & Jeremiah, A. (2023). Advanced statistical techniques in regression
analysis.
5. Bajari, P., Chu, C., & Park, M. (2012). "An Empirical Model of Subprime Mortgage Default from 2000
to 2007." Examines the use of machine learning models for predicting real estate market trends.
6. Byeonghwa, C. (2015). Comparative analysis of machine learning algorithms for housing price
prediction.
7. Galster, G., & Tatian, P. (2009). "Modeling Housing Appreciation Dynamics." Highlights the
application of statistical and machine learning techniques in housing market analysis.
8. Gao, Y., Chen, F., & Liu, Z. (2022). Enhancing prediction accuracy in housing price models with
advanced regression techniques.
9. Glaeser, E. L., Gyourko, J., & Saks, R. E. (2005). "Why Have Housing Prices Gone Up Analyzes the
drivers of housing price increases, focusing on supply and demand dynamics.
10. Gruber, M. H. J., & Schucany, W. R. (2020). Applications of ridge regression in econometrics,
chemistry, and engineering.
11. Hilt, D. E., & Seegrist, R. R. (1977). Ridge regression: A method of estimating the coefficients of
multiple regression models in scenarios where independent variables are highly correlated.
12. Ho, L. S., & Wong, G. (2008). Housing and macroeconomics: The role of asset prices in monetary
policy. Pacific Economic Review, 13(2), 223-239.
13. Hossain, M., & Latif, E. (2016). Housing price determinants in the context of macroeconomic
variables. Economic Modelling, 59, 174-183.
14. Jafari, M., & Akhavian, R. (2019). Square footage and its impact on house price prediction models.
15. Kain, J. F., & Quigley, J. M. (1970). "Measuring the Value of Housing Quality." Explores the
relationship between housing characteristics and prices using hedonic pricing models.
16. Khoo, K. L., Lee, C. H., & Goh, K. L. (2019). Housing prices and macroeconomic variables:
17. Paciorek, A. (2013). "Supply Constraints and Housing Market Dynamics." Discusses how supply
constraints influence housing prices and market behavior.
18. Raga Madhuri, G., Anuradha, G., & Vani Pujitha, K. (2019). Regression techniques for house price
prediction: A comparative study.
19. Temur, A., Akgün, A., & Temur, S. (2019). Gradient boosting and AdaBoost Regression for house
price prediction.
20. Yu, J., Liu, H., & Zhang, W. (2016). The influence of structural features on housing price prediction.