# RECORD 8a99113b012c7dc5288863734eba9456a85fd906
# TITLE: Investigating the impact of property characteristics, cost of living, and environmental factors on rental prices in Baidoa’s climate-affected real estate market: a hybrid approach using hedonic regression and neural networks
# YEAR: 2025.0  VENUE: Frontiers in Sustainable Cities  DOI: 10.3389/frsc.2025.1511761
# AUTHORS: Mohamed Ibrahim Nor; Buba Audu; A. Mohamed
# PAGES: 15  CANDIDATE TABLE PAGES (1-based): [7, 8, 9]
# ABSTRACT: Understanding the determinants of residential rental prices is crucial for policymakers, investors, and real estate practitioners. This study investigates the influence of property value, property characteristics, cost of living, political stability, essential services, and environmental factors on rental prices in Baidoa city. Additionally, the research compares different modeling approaches to enhance rental price forecasting.A dual-method approach was employed, integrating hedonic regression analysis and artificial neural network (ANN) models to analyze rental values. The dataset includes key variables such as the number of bedrooms, essential services, cost of living, and environmental conditions. The predictive performance and interpretability of both models were assessed to determine their effectiveness in rental price estimation.The findings reveal that rental prices are significantly influenced by the number of bedrooms, essential services (e.g., electricity), cost of living, and environmental conditions. However, political stability and displacement did not exhibit significant effects. While hedonic regression provided clear, interpretable insights into direct predictors, ANN models captured nonlinear interactions and demonstrated superior prediction accuracy. Nevertheless, the ANN model exhibited mixed performance, with 53% of cases underperforming and 47% exceeding predictions, highlighting the need for improved precision in forecasting.The study emphasizes the importance of a mixed-method approach in rental price forecasting. Policymakers should integrate econometric and machine learning models to refine housing policies and ensure fair market regulations. Investors and property owners can leverage these findings to optimize rental pricing strategies, while real estate practitioners can benefit from data-driven decision-making. This research contributes to the real estate valuation literature by bridging traditional econometric analysis with advanced machine learning techniques. The study validates the applicability of hedonic pricing and information asymmetry theories within an emerging market context, offering a more comprehensive understanding of rental price determinants.


===== PAGE 1 (score 7, numeric-paren 0, table-words 0, var-words 7) =====

Frontiers in Sustainable Cities
01
frontiersin.org
Investigating the impact of 
property characteristics, cost of 
living, and environmental factors 
on rental prices in Baidoa’s 
climate-affected real estate 
market: a hybrid approach using 
hedonic regression and neural 
networks
Mohamed Ibrahim Nor 1*, Buba Audu 2 and 
Abdullahi Dahir Mohamed 3
1 Institute of Climate and Environment (ICE), SIMAD University, Mogadishu, Somalia, 2 Buruuj 
Construction and Real Estate, Mogadishu, Somalia, 3 Faculty of Economics and Management Sciences, 
Hormuud University, Mogadishu, Somalia
Introduction: Understanding the determinants of residential rental prices 
is crucial for policymakers, investors, and real estate practitioners. This study 
investigates the influence of property value, property characteristics, cost of 
living, political stability, essential services, and environmental factors on rental 
prices in Baidoa city. Additionally, the research compares different modeling 
approaches to enhance rental price forecasting.
Methods: A dual-method approach was employed, integrating hedonic 
regression analysis and artificial neural network (ANN) models to analyze rental 
values. The dataset includes key variables such as the number of bedrooms, 
essential services, cost of living, and environmental conditions. The predictive 
performance and interpretability of both models were assessed to determine 
their effectiveness in rental price estimation.
Results: The findings reveal that rental prices are significantly influenced by 
the number of bedrooms, essential services (e.g., electricity), cost of living, 
and environmental conditions. However, political stability and displacement 
did not exhibit significant effects. While hedonic regression provided clear, 
interpretable insights into direct predictors, ANN models captured nonlinear 
interactions and demonstrated superior prediction accuracy. Nevertheless, the 
ANN model exhibited mixed performance, with 53% of cases underperforming 
and 47% exceeding predictions, highlighting the need for improved precision in 
forecasting.
Discussion: The study emphasizes the importance of a mixed-method approach 
in rental price forecasting. Policymakers should integrate econometric and 
machine learning models to refine housing policies and ensure fair market 
regulations. Investors and property owners can leverage these findings to 
optimize rental pricing strategies, while real estate practitioners can benefit 
from data-driven decision-making. This research contributes to the real estate 
valuation literature by bridging traditional econometric analysis with advanced 
machine learning techniques. The study validates the applicability of hedonic 
OPEN ACCESS
EDITED BY
Jianquan Cheng, 
Manchester Metropolitan University, 
United Kingdom
REVIEWED BY
Olatunji Shobande, 
University of Aberdeen, United Kingdom
Debora Anelli, 
Politecnico di Bari, Italy
*CORRESPONDENCE
Mohamed Ibrahim Nor 
 m.ibrahim@simad.edu.so
RECEIVED 15 October 2024
ACCEPTED 03 March 2025
PUBLISHED 19 March 2025
CITATION
Nor MI, Audu B and Mohamed AD (2025) 
Investigating the impact of property 
characteristics, cost of living, and 
environmental factors on rental prices in 
Baidoa’s climate-affected real estate market: 
a hybrid approach using hedonic regression 
and neural networks.
Front. Sustain. Cities 7:1511761.
doi: 10.3389/frsc.2025.1511761
COPYRIGHT
© 2025 Nor, Audu and Mohamed. This is an 
open-access article distributed under the 
terms of the Creative Commons Attribution 
License (CC BY). The use, distribution or 
reproduction in other forums is permitted, 
provided the original author(s) and the 
copyright owner(s) are credited and that the 
original publication in this journal is cited, in 
accordance with accepted academic 
practice. No use, distribution or reproduction 
is permitted which does not comply with 
these terms.
TYPE  Original Research
PUBLISHED  19 March 2025
DOI  10.3389/frsc.2025.1511761


===== PAGE 7 (score 44, numeric-paren 6, table-words 12, var-words 8) =====

Nor et al.
10.3389/frsc.2025.1511761
Frontiers in Sustainable Cities
07
frontiersin.org
learning algorithm with a suitable loss function to minimize the 
prediction error. An ANN’s functional relationship can be expressed 
as follows:
	
1
2 3
4
(
,
,
,
p
y
f x x x
x
x
=
……………
where 1x , 2
3
4
,
,
x
x
x ………………. p
x  are the input variables and y 
is a target variable.
The MLP algorithm relies heavily on hidden units, which compute 
the weighted sum of their inputs to obtain their scalar net activation. 
The net activation is the inner product of the inputs and the weights 
at the hidden unit, which is expressed as follows:
	
0
1
0
d
d
T
j
i
ji
i
ji
j
i
i
net
x w
w
x w
W X
=
=
=
+
=
≡
∑
∑
This illustrates that every hidden unit emits an output that 
varies according to how it is activated. Table 1 presents the results 
of the experiment, the inputs (predictors), and the sampling 
strategy. The training, test, and validation sets of the experimental 
data are used to obtain the best results. The sample size of each 
component of the data is shown in Table  1. Similarly, the 
architecture of the MLP algorithm is explained in Table 1 and 
Figure 1. An MLP with at least three hidden layers and a maximum 
of ten layers is used in this investigation. In addition, five networks 
are 
retained, 
while 
20 
networks 
are 
used 
for 
the 
training experiment.
Results of hedonic regression
The results of the hedonic regression analysis indicate that 
property features, access to essential services, economic factors, 
political conditions, and displacement have significant impacts on 
rental property rates. Our model is robust because it explains a 
significant proportion of the variance in rental values (R2 = 0.96, 
F = 105.46, p < 0.000). It has also been established that the model has 
TABLE 2  Hedonic model estimates.
Independent variables
Coefficient
t-statistic
VIF
Cohens f2
Constant
143.58
4.97
Residential property price (RPP)
0.02
0.61
1.15
9.07
Number of bedrooms (NBR)
0.09
2.72***
1.4
9.15
Dedicated parking space (DPS)
−0.01
−0.36
1.64
9.07
Residential property size (RPS)
0.01
0.26
1.24
9.15
Access to electricity (AEL)
−0.18
−5.96***
1.18
9.09
Environmental factors
−0.12
−3.71***
1.39
9.14
Cost of living (COL)
−0.26
−4.85***
3.65
9.04
Political stability (POS)
−0.02
−0.66
1.41
9.13
Displacement (DIS)
0.05
1.26
1.68
9.11
Interaction1 (I1)
0.25
4.49***
3.93
8.86
Interaction2 (I2)
0.67
12.24***
3.82
8.85
R-squared (R2)
0.96
Adjusted (Adj. R2)
0.92
F-statistics (prob)
105.46
0.00
Durbin-Watson test
2.05
0.965
Heteroscedasticity test
No heteroscedasticity concerns were found in the data.
Normality test: Kolmogorov–Smirnov
0.07
0.596
Mean squared error (MSE)
1507.13
Mean absolute error (MAE)
29.38
Dependent variable: Residential property monthly rentals (RPMR).
**p < 0.05; ***p < 0.01.
TABLE 3  ANN test results (forecasts).
Model/item
Number of 
items
% of items
Item: less 
than 10%
% of items 
(within 10%)
Item: more 
than 10%
% of items 
(more than 
10%)
Over-performed
7
23.33
2
28.57
5
71.43
Under-performed
23
76.67
3
13.04
20
86.96
Total
30
100
5
100
25
100


===== PAGE 8 (score 19, numeric-paren 0, table-words 5, var-words 7) =====

Nor et al.
10.3389/frsc.2025.1511761
Frontiers in Sustainable Cities
08
frontiersin.org
high explanatory power, indicating the significant influence of the 
independent variables on monthly rental values.
The underlying assumptions of the regression model were 
carefully tested to confirm the robustness of the findings. The residuals 
of the model were normally distributed, as verified by the 
Kolmogorov–Smirnov test (stat. = 0.07; p-value = 0596), confirming 
that the assumption of a normal distribution was satisfied. 
Furthermore, the Durbin-Watson test is employed to ensure the 
absence of significant autocorrelation among residuals, and as 
reported in Table 2, the Durbin-Watson test (DW = 2.45, p = 0.053) 
affirmed the independence of the residuals of the model. While 
checking for multicollinearity, we found that all variance inflation 
factor (VIF) values are below five (5), signifying that there is no 
considerable redundancy among the predictors. Moreover, as reported 
in Table 2, the Cohen’s f2 test indicates a strong effect of the predictors 
on the dependent variable, well above the established benchmarks. 
These tests validate the integrity of the analysis, highlighting the 
reliability of the model.
Remarkably, the study revealed that the number of bedrooms in a 
property has a direct and positive effect on property rental prices 
(b = 0.09, t = 2.72, p < 0.00), implying that larger homes demand 
higher rents. In addition, properties with greater area demand an 
increase in rental prices (b = 0.07, t = 2.26, p < 0.05), confirming that 
spatial dimensions play a critical role in rental valuation. In contrast, 
the study reveals significant negative effects from socioeconomic 
factors such as high cost of living (b = −0.26, t = −4.85, p < 0.01) and 
political instability (b = −0.08, t = −2.48, p < 0.05), which reduce 
property rental prices, exemplifying the burden of economic strain on 
rental affordability, political risks and uncertainties, which might 
prevent long-term residential commitments or weaken consumer 
confidence. These findings underscore the sensitivity of property 
rental markets to economic and political climates, showing the 
necessity for good governance and effective economic policies to 
sustain healthier rental markets.
Shockingly, despite Baidoa city’s vulnerability to ecological shocks, 
the analysis indicates that environmental factors such as significant 
climate change and sustainable practices do not significantly affect 
rental decisions. These results may reflect a gap in tenant awareness or 
prioritization regarding environmental sustainability and resilience in 
the residential property market. Moreover, the study revealed a 
positive relationship between the influx of internally displaced persons 
(IDPs) and property rental prices (b  = 0.07, t  = 2.18, p  < 0.05), 
demonstrating that demographic pressures from displacement can 
aggravate housing demand, thus increasing property rental prices in 
affected regions. The interaction terms further illuminate the 
compound effects of combined variables on property rental prices. 
Remarkably, the interaction between property rental rates and the cost 
of living significantly increases rental rates (b = 0.21, t = 3.52, p < 0.01), 
FIGURE 1
ANN model.
TABLE 4  Comparison metrics.
Model
Mean squared error (MSE)
Mean absolute error (MAE)
Regression model
1507.13
29
ANN model
6920.83
65.75


===== PAGE 9 (score 20, numeric-paren 0, table-words 4, var-words 9) =====

Nor et al.
10.3389/frsc.2025.1511761
Frontiers in Sustainable Cities
09
frontiersin.org
whereas the synergy between property rental rates and essential 
services, principally electricity, is even more noticeable (b = 0.74, 
t = 12.48, p < 0.01). These results suggest that while individual factors 
influence property rental prices, their interdependencies can intensify 
impacts, underscoring the complexity of rental market dynamics. This 
comprehensive analysis assists policymakers and stakeholders in 
understanding the multidimensional effects on property rental prices, 
providing a basis for informed decision-making intended to 
accomplish sustainable urban development and housing affordability.
The results of the hedonic regression analysis indicate that several 
factors significantly influence residential property prices. Our model 
demonstrates robust explanatory power, accounting for a substantial 
proportion of the variance in property values (R2 = 0.96, F = 105.46, 
p < 0.00), signaling the significant impact of the independent variables 
on monthly rentals of residual properties. The underlying assumptions 
of the regression model were thoroughly assessed to ensure the 
reliability of the results. The residuals of the model were found to 
be normally distributed, as confirmed by the Kolmogorov–Smirnov 
test (stat. = 0.07, p = 0.596), ensuring the normality assumption was 
met. Additionally, the Durbin-Watson test (DW = 2.05, p = 0.965) 
confirmed the absence of autocorrelation among the residuals, 
affirming the independence of the error terms. Regarding 
multicollinearity, all variance inflation factor (VIF) values were below 
five (<5), suggesting no significant redundancy among the predictors. 
The Cohen’s f2 values further indicate a substantial effect of the 
predictors on the dependent variable, underscoring the robustness of 
the model. Among the significant variables, the number of bedrooms 
(b  = 0.09, t  = 2.72, p  < 0.01) shows a positive relationship with 
property prices, suggesting that properties with more bedrooms tend 
to have higher values. The cost of living (b = −0.26, t = −4.85, p < 0.01) 
and access to electricity (b = −0.18, t = −5.96, p < 0.01) were found to 
have a significant negative impact on property prices, indicating that 
higher costs of living and limited access to electricity contribute to 
lower property values. The environmental factors (b  = −0.12, 
t = −3.71, p < 0.01) also exert a significant negative effect, suggesting 
that environmental challenges, such as climate change, negatively 
impact property prices. Furthermore, the interaction between 
monthly rentals of residual properties and the cost of living (b = 0.25, 
t = 4.49, p < 0.01) and the interaction between monthly rentals of 
residual properties and access to electricity (b  = 0.67, t  = 12.24, 
p < 0.01) were both found to significantly amplify monthly rentals of 
residual properties. These interaction terms emphasize how the 
combination of economic factors and essential services can intensify 
their effects on property prices. These findings highlight the key 
drivers of residential monthly rentals, with economic and service-
related factors playing a critical role. Policymakers and urban planners 
should consider these insights to develop effective housing strategies 
that address the complexities of property price dynamics.
Results of the artificial neural network
In this study, we employed an artificial neural network (ANN) 
technique to forecast property rental values for Baidoa city using a 
multilayer perceptron (MLP) model. In this study, data preprocessing 
was conducted to ensure the integrity and usability of the dataset 
before training the artificial neural network (ANN) model. The dataset 
was partitioned into three subsets: a training set comprising 70 
samples (60%), a testing set with 30 samples (25%), and a validation 
set containing 18 samples (15%). A total of 11 predictors were 
identified (from the regression results), with numerical features 
standardized to ensure uniformity and categorical variables encoded 
to facilitate machine learning processing. The ANN model was 
structured with an input layer consisting of 11 nodes, corresponding 
to the number of predictors, and a hidden layer architecture ranging 
between 3 and 10 layers, with the final optimized model incorporating 
five hidden layers. The ReLU activation function was applied to the 
hidden layers to introduce non-linearity, while a linear activation 
function was used in the output layer to predict the Monthly Property 
Rentals (MPR). The Adam optimizer was employed to adjust model 
weights efficiently, and the mean squared error (MSE) loss function 
was used to measure prediction accuracy. The model underwent 2000 
training iterations, ultimately achieving an MSE of 6920.83 and a 
mean absolute error (MAE) of 65.75, indicating an optimal predictive 
performance (as illustrated in Table 3).
A significant valuation difference is evident in Figure 1, indicating 
significant fluctuations in prediction accuracy, with both 
underestimations and overestimations evident. Thirty property cases 
were critically evaluated to determine the effectiveness of the ANN 
model for predicting Monthly Property Rentals (MPR). The evaluation 
yielded two primary outcomes regarding the model’s performance: 
underperformance and over-performance against actual rental values. 
Among the 30 cases, 16 properties (53%) exhibited some degree of 
underperformance relative to their actual rental values, while 14 cases 
(47%) demonstrated over-performance (see Table 4). The data indicate 
a higher occurrence of underperformance in the predictions, where 
the majority (71%, or 11 out of 16 cases) underestimated rental prices 
by less than 10%, whereas the remaining five (5) cases exceeded this 
threshold, suggesting that moderate underestimation existed at 31%. 
Specifically, 16 properties—accounting for over half of the evaluated 
samples—failed to meet their actual rental values, signaling potential 
revenue discrepancies for property owners and investors. The 
predominance of underperformance, mostly within a 10% margin, 
reflects a conservative estimation pattern. While this approach 
mitigates the risk of rental overvaluation that could lead to unrealistic 
investment expectations, it simultaneously risks undervaluing 
properties, potentially limiting revenue generation opportunities and 
affecting the overall financial viability of rental investments. The 
findings from the ANN model evaluation highlight critical insights for 
stakeholders in the rental market, indicating both opportunities and 
challenges associated with current predictive modeling approaches. 
These insights inform strategic decision-making in pricing, 
forecasting, investment, and market entry strategies.
Pricing strategies adjustment
The analysis reveals that a significant proportion (53%) of the 
forecasts were lower than actual market rental values, indicating a 
persistent bias toward conservative pricing strategies among property 
owners and managers. While this conservative approach helps 
mitigate the risk of high vacancy rates, it also results in lower 
achievable rental earnings. To better align with actual market 
conditions, property managers should consider refining their pricing 
strategies, ensuring they accurately reflect market demand while 
maintaining competitiveness. Overly cautious pricing may limit 
potential revenue streams, requiring a more dynamic and data-driven 
approach to rental valuation.


===== PAGE 10 (score 8, numeric-paren 0, table-words 2, var-words 4) =====

Nor et al.
10.3389/frsc.2025.1511761
Frontiers in Sustainable Cities
10
frontiersin.org
Development of forecast models
The substantial deviations observed between forecasted and actual 
rental values suggest that further refinements to predictive modeling 
are necessary. In some cases, rental values were significantly 
overestimated, while in others, they were considerably underestimated. 
These inconsistencies emphasize the need for incorporating a broader 
set of explanatory variables, including economic indicators, 
demographic trends, and historical market fluctuations, to enhance 
forecast accuracy. Developers should integrate real-time economic 
data reflective of the market’s conditions to improve model precision 
and bridge the gap between predicted and actual values.
Opportunities for niche investments
The findings indicate significant investment opportunities within 
specific market segments. Properties that consistently outperform 
their forecasts, particularly those exceeding predicted values by more 
than 10%, are likely undervalued and may provide higher-than-
expected returns. Investors should consider these properties as high-
potential opportunities for maximizing rental income. Conversely, 
areas where rental forecasts frequently underperform may require 
additional due diligence, as they could signal market volatility or 
external factors affecting rental stability. Identifying these trends 
enables investors to allocate capital strategically and optimize returns.
Market entry strategies and portfolio 
management
The mixed accuracy of forecasts suggests that prospective investors 
should adopt a diversified approach to portfolio management. While 
some properties align closely with predicted rental values, others exhibit 
substantial variation, either exceeding or falling short of expectations. To 
mitigate risks, investors should consider a balanced portfolio, including 
stable properties with predictable returns and speculative investments 
with higher growth potential. This strategic diversification allows market 
participants to capitalize on emerging opportunities while safeguarding 
against unforeseen market fluctuations.
Advanced market analysis and continual 
monitoring
The variability in rental market forecasts underscores the 
importance of ongoing market analysis and continuous model 
adjustments. Given the dynamic nature of the housing market, real 
estate professionals must remain vigilant and responsive to economic 
and regulatory changes that may impact rental demand. Regular 
updates to predictive models, informed by market trends and real-
time data, will enhance forecasting accuracy and enable property 
owners to make proactive adjustments. By continually refining their 
strategies, stakeholders can maintain a competitive edge and ensure 
long-term financial sustainability in the rental market (Figure 2).
Comparison of regression and ANN model 
performance
The regression model outperforms the artificial neural network 
(ANN) model in terms of predictive accuracy, as indicated by its 
significantly lower error metrics. The regression model achieves a 
Mean Squared Error (MSE) of 1507.13 and a Mean Absolute Error 
(MAE) of 29, suggesting that its predictions are closer to the actual 
values. In contrast, the ANN model exhibits a much higher MSE of 
6920.83 and an MAE of 65.75, indicating greater deviations from the 
true values (see Table 5). This suggests that, in this particular case, the 
regression model is more effective in minimizing errors and providing 
more reliable predictions compared to the ANN model.
Discussion
The main aim of this study is to investigate the influence of 
property value, property characteristics, cost of living, political 
stability, essential services and environmental factors on residential 
property rental prices in Baidoa city. This study uses a dual-method 
TABLE 5  Summary of data preprocessing, ANN model configuration, and model performance results.
Stage
Details
Data preprocessing
Training set: 70 samples (60%)
Testing set: 30 samples (25%)
Validation set: 18 samples (15%)
Feature count: 11 predictors
Normalization: standardized numerical predictors
Categorical encoding: converted non-numeric predictors
ANN (MLP) model
Input layer: 11 nodes (one for each predictor)
Hidden layers: between 3 and 10 layers (based on tuning)
Output layer: 1 node (predicting MPR)
Activation functions:
- ReLU for hidden layers
- Linear for output
Optimizer: Adam
Loss function: mean squared error (MSE)
ANN model results
Mean squared error (MSE): 6920.83
Mean absolute error (MAE): 65.75
Training iterations: 2000
Hidden layers: 5