The present work aims to improve the design efficiency and optimize the results in the increasingly complex and diversified material design projects to help architects realize the better performance of building structures. According to the characteristics of comprehensive perception and intelligent processing of the Internet of Things, a reverse suspension structure design model is constructed based on the finite element method and simulated annealing algorithm. Besides, deep learning is adopted to train complex functions for performance correction and to optimize the plane structure of shell structure. Moreover, the force is transformed into shape, and the form-finding process is completed to facilitate the operation of designers. Finally, the spatial anchoring ability of the geographic information system is used to match and calculate the relevant characteristics of spatial elements. On this basis, the index construction strategy based on weight distribution is employed to realize the data fusion diagnosis framework and enhance the intelligence of architectural design. The simulation results show that the maximum tensile stress of the physical suspension experiment is 3.71 MPa and the maximum compressive stress is 14.7 MPa. The compressive stress value is much larger than the tensile stress value. The maximum deformation value’s difference between the compressive and tensile stress is 0.07 and 0.11, respectively. The error is within the acceptable range, which is similar to the compression state results obtained from the actual suspension physical experiment, indicating that the initial design model of the reverse suspension structure based on deep learning is reliable. In addition, the evolutionary optimization effect analysis results demonstrate that the load of the design structure is relatively uniform, which verifies the feasibility of the algorithm reported here. The research significance of the reverse suspension structure model constructed here is to provide an accurate and feasible design idea for the reverse design of some complex structures in the building suspension. It can also shorten the creation and improvement cycle of this kind of structure and optimize the performance and construction cycle of the building structure.

The present work aims to improve the comfort of architectural interior design and reduce indoor energy consumption. The Weight K-Nearest Neighborhood (WKNN) algorithm and Nondominated Sorting Genetic algorithm are proposed to locate and analyze the spatial location of indoor personnel and optimize the indoor energy consumption in combination with residential behavior. Firstly, the indoor human behavior data and energy-saving problems are analyzed based on residential behavior theory and architectural physics. The indoor positioning algorithm is proposed to identify the personnel activities to realize the optimization of indoor energy distribution. Secondly, mean filtering and cluster analysis are adopted to optimize sampling points’ data and fingerprint database to eliminate data noise. Besides, the WKNN algorithm is used for Wireless Fidelity (Wi-Fi) indoor location fingerprint location. Then, aiming at the multiobjective optimization problem of building indoor energy consumption, the Nondominated Sorting Genetic algorithm obtains the optimal solution of the model. Combined with the indoor location information of personnel, the indoor heating and cooling system is monitored and distributed to reduce the energy consumption in the building while ensuring the living comfort of personnel. The test and simulation results demonstrate that the mean filtering algorithm can solve the room’s fluctuation problem of Wi-Fi signals. The cluster analysis algorithm can eliminate the data noise of the fingerprint database and improve the positioning accuracy of the positioning algorithm. Moreover, the location result is independent of the number of nodes; the number of sampling points does not affect the location result. The positioning accuracy of the WKNN algorithm is 2 m, and the positioning error rate within 2 m is about 60%. Compared with other algorithms, the WKNN algorithm has better positioning accuracy and positioning stability. Therefore, the location algorithm designed here can be applied to indoor location optimization. This study provides a reference for optimizing buildings’ indoor positioning and energy consumption.