# RECORD db3684d67741de25ae92064ef093bf9c1d604ef8 # TITLE: Analyzing Key Determinants of Property Valuation in Boston Neighborhoods # YEAR: 2025.0 VENUE: Advances in Economics, Management and Political Sciences DOI: 10.54254/2754-1169/2025.bj25752 # AUTHORS: Shuo Xin # PAGES: 9 CANDIDATE TABLE PAGES (1-based): [6] # ABSTRACT: The current research examines the effects of particular attributes of dwellings on property valuation in neighborhoods of Boston. Given the increased sophistication of urban property markets, particularly in historically dense cities such as Boston, this study aims to offer data-based insights into factors most influential in property valuation. Employing the Boston 2024 property assessment data set from the Boston open data portal, this research utilizes multiple linear regression and K-means cluster analysis to conduct a broad investigation. Multiple linear regression measures the relationship between physical attributesland value, number of bathrooms, and living spaceand appraisal value. In contrast, K-means clustering identifies latent property groupings involving similarities in structure and market value level. Results demonstrate that area size, number of full bathrooms, and land value are the main contributing factors in higher values, while other factors, such as number of kitchens and some ZIP codes, negatively impact property value. These findings offer actionable information to homeowners, builders, and investment, urban design, and housing policy in Boston. ===== PAGE 1 (score 12, numeric-paren 0, table-words 0, var-words 12) ===== Proceedings of ICEMGD 2025 Symposium: The 4th International Conference on Applied Economics and Policy Studies DOI: 10.54254/2754-1169/2025.BJ25752 © 2025 The Authors. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0 (https://creativecommons.org/licenses/by/4.0/). 196 Analyzing Key Determinants of Property Valuation in Boston Neighborhoods Shuo Xin D'Amore-McKim School of Business, Northeastern University, Boston, USA shuoxin4@gmail.com Abstract. The current research examines the effects of particular attributes of dwellings on property valuation in neighborhoods of Boston. Given the increased sophistication of urban property markets, particularly in historically dense cities such as Boston, this study aims to offer data-based insights into factors most influential in property valuation. Employing the Boston 2024 property assessment data set from the Boston open data portal, this research utilizes multiple linear regression and K-means cluster analysis to conduct a broad investigation. Multiple linear regression measures the relationship between physical attributes—land value, number of bathrooms, and living space—and appraisal value. In contrast, K-means clustering identifies latent property groupings involving similarities in structure and market value level. Results demonstrate that area size, number of full bathrooms, and land value are the main contributing factors in higher values, while other factors, such as number of kitchens and some ZIP codes, negatively impact property value. These findings offer actionable information to homeowners, builders, and investment, urban design, and housing policy in Boston. Keywords: Property Valuation, Linear Regression, K-means Clustering 1. Introduction American city real estate markets are volatile, with housing affordability and urban planning as key issues. Boston's housing market is unique, with heterogeneous neighborhoods and fluctuating property values [1]. In the context of expanding urban populations and real estate investment, it is critical for stakeholders to understand the factors that influence property appraisal values [2]. Prior research has established the impact of urban amenities, public transit, and educational infrastructure on property valuations. Prior research offers broad housing price insights, but neglects Boston's micro-market dynamics. Existing models often lack interpretability and local feature integration. This study aims to fill that gap by investigating how specific physical features of properties—such as living area, number of bathrooms, year of remodeling, land value, and more— correlate with and influence property appraisal values in different neighborhoods across Boston. This study utilizes multiple linear regression and K-means clustering to analyze the relationship between housing characteristics and property appraisal value. This study selects explicit regression because the prediction target—appraisal value—is a continuous, numeric variable, making regression analysis more appropriate than classification methods designed for categorical or ===== PAGE 6 (score 14, numeric-paren 0, table-words 2, var-words 8) ===== Proceedings of ICEMGD 2025 Symposium: The 4th International Conference on Applied Economics and Policy Studies DOI: 10.54254/2754-1169/2025.BJ25752 201 (RMSE) was roughly $279,830, which reflects the typical deviation of the model's predictions from actual property values. 2.5. K-means clustering analysis K-means clustering is an unsupervised learning algorithm. It divides the data into K non- overlapping clusters by calculating the distances between samples [6]. K-means clustering was applied as an unsupervised learning technique to uncover hidden patterns in the property dataset beyond supervised modeling. A diverse sample of 7,000 properties was used to reduce computational load. The Elbow Method and Silhouette Scores were used to determine the optimal number of clusters, with the Elbow plot indicating three clusters. The Silhouette analysis showed relatively high scores for three clusters, suggesting that this choice would yield well-separated and cohesive groupings. Figure 5. Elbow method curve and silhouette score plot for K-means Based on the clustering results: •  Cluster 0 predominantly included larger family homes with substantial gross-living areas and moderate appraisal values. • Cluster 1 consisted of smaller, lower-valued properties, often characterized by limited living space and fewer amenities. •  Cluster 2 represented luxury homes with expansive living spaces, multiple bathrooms, and the highest overall appraisal values. Table 2. Mean feature values across K-means clusters Clust er LIVING_A REA FULL_ BTH LAND_VA LUE HLF_B TH BED_R MS KITCH ENS GROSS- LIVING LAND_ SF OWN_OC C_Y TOTAL_V ALUE 0 2356.8336 77 2.37571 6 198065.57 85 0.3762 89 4.30870 6 1.66666 7 1005.19553 3 3471.576 174 0.56071 1240936 1 1212.7580 6 1.3954 57066.510 8 0.173 2.381 1.1462 371.98534 1909.302 2 0.5344 591606.8 2 2787.5183 07 2.74409 4 354708.66 14 0.7401 57 3.83070 9 1.33070 9 517.245472 2948.070 866 0.614173 2929574 These clusters(Table 2) provide a valuable segmentation of the Boston real estate market, complementing the regression findings and offering additional perspectives for understanding ===== PAGE 7 (score 19, numeric-paren 0, table-words 4, var-words 11) ===== Proceedings of ICEMGD 2025 Symposium: The 4th International Conference on Applied Economics and Policy Studies DOI: 10.54254/2754-1169/2025.BJ25752 202 property value dynamics across different neighborhoods. 3. Results and discussion 3.1. Correlation analysis results Correlation Analysis is a statistical method used to quantify the strength and direction of the linear relationship between two or more variables, and its results are usually expressed by the correlation coefficient (between -1 and 1) [7]. The correlation analysis provided initial insights into the relationships between various physical features of properties and their appraisal values. As shown in Figure 1, the correlation heatmap revealed that the living area exhibited the strongest positive relationship with property value among all features considered. The results align with real estate market trends, where larger living areas correlate with higher prices. Full bathrooms and land value also showed positive correlations, indicating their significant contribution to property valuation. Conversely, specific features, such as the number of kitchens and the number of bedrooms, showed weaker or even slightly negative correlations with property value. This result implies that beyond a certain threshold, adding more kitchens or bedrooms may not linearly increase a property's market value and may even indicate property types with lower per-unit value, such as multi-family residences. The bar plot of feature correlations presented in Figure 2 further ranked these relationships quantitatively. Living area and full bathrooms emerged as dominant contributors, confirming their importance for further modeling steps. Additionally, the boxplot in Figure 3 illustrates that owner- occupied properties tend to have slightly higher median appraisal values than non-owner-occupied properties. This observation suggests that owner occupancy could be a proxy for better property maintenance and investment in amenities, ultimately enhancing market value. 3.2. Multiple linear regression model findings The multiple linear regression model provided interpretable results, quantitatively linking property characteristics to appraisal values [8]. Regression coefficients indicated that an additional square foot of living area increased appraisal value by approximately $414, with additional full bathrooms also contributing positively to valuation. Land value maintained a steady positive slope. Interestingly, the number of kitchens had a negative coefficient, indicating that properties with multiple kitchens, possibly multi-unit dwellings, tended to have lower appraisal values when controlling for other factors. This outcome emphasizes that not all amenities uniformly increase property worth. A critical dimension uncovered through the regression analysis was the effect of ZIP codes on property values. Figure 5 displays the ZIP code coefficient distribution, illustrating Boston's significant geographic disparities. ZIP code 02108 (Beacon Hill) properties had the highest positive adjustment of over $589,000, reflecting the area's historical prestige, central location, and architectural significance. In contrast, ZIP code 02121 (Dorchester) was associated with a substantial negative adjustment of approximately $432,000. This contrast highlights how location, independent of structural features, is critical in determining real estate market dynamics in urban environments. The model achieved an R-squared value of 0.736, indicating a strong fit, with approximately 74% of the variation in property appraisal values explained by the selected features. The Root Mean Square Error (RMSE) of $279,830 is reasonable given Boston's overall high property price levels.