Smart home management based on deep learning: Optimizing device prediction and user interface interaction

This work aims to address the challenges faced by smart home systems, including the accuracy of device status prediction, user interface design, system stability, and response speed. As smart home devices become more widely used, the need for accurate predictions of their operational status has increased. This includes predicting the switch states, faults, and performance metrics of devices such as smart lights, thermostats, and security systems. To address this demand, an innovative multimodal prediction model combining the Convolutional Neural Network and Long Short-Term Memory network is proposed to enhance the accuracy of smart device status predictions. Cloud computing technology is used for the user interface design to create an intuitive and user-friendly interface, ensuring both system stability and fast response times. The experiments compare the performance of the proposed model with traditional models in predicting the status of smart devices. The results demonstrate that the proposed system reduces the Mean Squared Error and Mean Absolute Error by 20% and 15%, respectively, significantly improving prediction performance. Furthermore, user satisfaction surveys indicate a 25% increase in satisfaction with the system. The proposed system also reduces the utilization rates of the Central Processing Unit, memory, Graphics Processing Unit, and network bandwidth by 15%, 18%, 25%, and 20%, respectively. These findings highlight the system's advantages in accuracy, user satisfaction, and resource utilization efficiency, providing strong support for the design and application of smart home systems.