Hesitant Fuzzy Long Short-Term Memory Network and Its Application in the Intelligent Building Selection

As the deep learning algorithm, the long short-term memory (LSTM) network is an emerging and hot tool to address classification issues. In the classification process, to provide more reasonable suggestions, the subjective evaluation and qualitative selection given by decision makers and experts are important and cannot be ignored. Thus, this paper proposes the hesitant fuzzy LSTM (HF-LSTM) network model to simultaneously address these two issues. To do this, we first define the series-connection hesitant fuzzy set (SHFS) and the classified SHFS to show the internal logic and the sequential order among the qualitative data. They could be more suitable presentation tools to present subjective evaluation information than other fuzzy sets. Concerning the new characters and properties of these new fuzzy sets, the HF-LSTM network is further constructed; also, the generalized and dispersed properties are proven mathematically. Then, the HF-LSTM network model is suitable for these new fuzzy environments and can provide classification results so that subjective evaluation and qualitative selection can be achieved in the whole classification process. Further, this paper derives the global optimization algorithm of the HF-LSTM network which can make the final classification and improve the credibility of the classification results. Finally, we apply the proposed methods to intelligent building selection and the results present the feasibility of the HF-LSTM network and its global optimization.