A Holistic Framework for Building Type Classification Using CNN and Random Forest
Ricky Rajora, Deepak Banerjee, Rahul Singh Chauhan, Hemant Singh Pokhariya, Gotte Ranjith Kumar · 2024
This study has developed a cutting-edge CNN model to classify various building types. This model has the potential to revolutionize urban management and planning. The results presented in Table 1 reveal the model's impressive performance in categorizing buildings into three classes: Apartment, Industrial, and Other. Its precision, recall, and F1-score values consistently exceed 95%, indicating its exceptional ability to recognize and distinguish intricate patterns in building characteristics for accurate classification. This model offers a powerful tool for urban planning and management, significantly enhancing decision-making in various urban development aspects. The Macro Average, Weighted Average, and Micro Average values of the model demonstrate its consistency and adaptability. The model performs well with datasets of different compositions due to support proportions aligning with the number of instances in each class. The model's overall accuracy of 97% indicates its ability to make correct predictions. This makes it useful in urban resource allocation and disaster response planning. The Weighted Average values also highlight the model's ability to handle imbalances in class sizes, making it more reliable for practical applications. The Micro Average approach provides an overall measure with a focus on the model's overall performance. It's a powerful tool for classifying building types. The CNN-based approach is a big step forward in making building-type classification automatic. It is a highly accurate and efficient solution that can be used by cities and others to manage resources, plan for emergencies, and make decisions. The model's high performance across different measurements shows that it is a reliable tool for dealing with the complexity of categorizing different urban areas.