The housing information system involves personal sensitive data, and the user’s emotional state is crucial to the system interaction experience. This study constructs a housing information system user emotion recognition model based on the random forest algorithm, and explores the problems of EEG signal feature extraction and emotion classification. The study adopts the DEAP dataset and performs feature dimensionality reduction by two types of feature extraction methods, namely power spectral density and differential entropy, combined with SavitzkyGolay feature smoothing and minimum redundancy maximum correlation algorithms. The experiments set different emotion label thresholds on two emotion dimensions, arousal and validity, and compared the emotion recognition effects of decision tree and random forest algorithms. The results show that the classification accuracy of the random forest algorithm in the arousal and valence dimensions reaches 92.2% and 91.0%, respectively, which is much higher than that of the multilayer perceptual machine and close to the discrete emotion model. Compared with the three classifiers, LSTM, KNN and LR, the Random Forest classifier has an average training time of only 522 ms and a test prediction time of 29.6 ms, which is the best overall performance. The study confirms that the random forest algorithm is computationally efficient and resistant to overfitting when processing high-dimensional EEG signal data, and provides feasible technical support for emotion recognition for the optimization of user experience in housing information systems.
