AI-Driven Property Management Decision Support System Using LSTM Networks for Energy Optimization

Abstract

Background: IoT devices and sensors collect real-time data, which is used in conjunction with Long ShortTerm Memory (LSTM) models to predict energy consumption trends, optimize resource allocation, and analyze household feedback for service improvements. Aim: This study aims to develop an AI-powered property management decision support system to enhance management efficiency and service quality. Method: The research involves multiple stages, including data collection from IoT devices, data cleaning and preprocessing, construction of predictive models (such as LSTM networks), and system optimization. Performance is evaluated using key metrics like task completion rate, maintenance request resolution rate, and resident satisfaction. Cross-validation and model performance measures, such as Mean Absolute Error (MAE) and Mean Squared Error (MSE), are used to assess model accuracy. Results: Empirical results show significant improvements in property management metrics: task completion rate increased by 15%, maintenance request resolution rate improved by 10%, and resident satisfaction rose by 7%. The LSTM model achieved a prediction accuracy with an MAE of 12.5 and an MSE of 256.4. Staff efficiency also increased by 10%, demonstrating the system’s practical value in optimizing resource allocation and predictive maintenance. Conclusion: This study introduces a novel technical approach and practical guidance for intelligent property management, providing significant theoretical and practical implications for enhancing operational efficiency, service quality, and decision-making capabilities in the property management industry.