Analysis Method of User Experience Influencing Factors Based on Improved LSTM Algorithm Technology

Abstract

This article proposes a user experience influencing factor analysis method based on the improved LSTM (Long Short-Term Memory) algorithm. LSTM, as a special type of recurrent neural network, performs well in sequence data processing, but its performance may be limited when dealing with long sequences or complex user behaviors. Therefore, improvements were made to LSTM to enhance the performance of User Experience (UE) prediction. Firstly, a heterogeneous network containing users and words was constructed, and joint research on user and word representations was conducted by embedding network nodes. This method enables the obtaining of user and word representations with certain emotional polarity tendencies, providing strong support for subsequent emotional analysis. Secondly, to further improve the LSTM model's ability to capture key text features, a self-attention mechanism was introduced. Through the self-attention mechanism, the model can automatically learn the text features that have the greatest impact on the final emotional analysis and assign them higher weights. This mechanism enables the model to more accurately grasp the true emotions of users when dealing with long sequences or complex user behaviors. Finally, all weights are assigned to the corresponding feature vectors to obtain the text vector. These text vectors contain rich user emotional information and behavioral features, which can be used for subsequent UE prediction. The experimental outcomes show that the improved LSTM algorithm proposed in this paper performs significantly better than traditional LSTM models and bidirectional LSTM models in UE prediction. The productivity of LSTM+attention mode has been significantly improved. The accuracy of this model is 6.1% higher than LSTM and 7.4% higher than bidirectional LSTM mode. The LSTM+Attention model proposed in this article effectively improves the UE prediction performance of the LSTM model. Adding user behavior features to the mode can effectively improve the accuracy of UE prediction.