The Exploration of a Multi-Level Fuzzy Comprehensive Evaluation Model

Based on the exploration of a multilevel fuzzy comprehensive evaluation, this paper investigates the application of the principles of fuzzy mathematics and combines the Analytic Hierarchy Process (AHP) with fuzzy theory to conduct multilevel fuzzy comprehensive assessments. During the study, Various data processing tools, such as Excel, MATLAB, and SPSS, are employed. After initial preprocessing and selection, mathematical models are developed through multidimensional processing and statistical analysis. By integrating the Isolation Forest algorithm with the ARIMA model, abnormal values in the collected data are successfully eliminated. It also avoids the possibility of erroneously deleting normal data by the Isolation Forest model. Taking into account the time series effects among the potential variables, LSTM time series prediction model is ultimately chosen, mitigating the drawbacks of traditional RNN neural networks, such as gradient vanishing and exploding. By applying the model corresponding to the actual problem, the multilevel fuzzy comprehensive evaluation is discussed.