Application Study Of Interpretable Model In Tourism Volume Forecasting

With the rapid development of the tourism industry,the demand for using computer technology,especially artificial intelligence algorithm to build intelligent tourism is increasing.As an important research topic of tourism industry,tourism demand forecasting has been paid continuous attention to.Most of the existing tourism demand forecasting models focus on improving the prediction accuracy while ignoring the research on the interpretation of the prediction results,which leads to the lack of users’ trust in the model and low utilization rate.Therefore,this paper studies and builds a deep learning-based tourism interpretable demand forecasting model on the basis of existing interpretable models.In order to improve the explanatory ability of tourism volume prediction,this paper optimizes the explanatory ability of tourism volume in two stages: sample data selection and prediction model construction.In the selection stage of sample data of tourism volume,the network search intensity of each keyword crawled from the network is taken as the feature that affects the tourism demand.Firstly,the missing value and data balance operation are carried out on the data to ensure the availability and integrity of the data,and then the four transformations are used to construct new features.With Light Gradient Boosting Machine(Light Gradient Boosting Machine),feature selection can be performed to improve the interpretability of sample data sources.In the prediction model construction stage,the two-stage convolutional neural network structure(CNN)was adopted,and the Gradient-CAM method(Gradient-weighted Class Activation Mapping)was used to generate the attribute map.Based on Local Interpretable model-agnostic Explanation,the interpretability of the Model was further improved,and an evaluation standard for the Interpretable Model was established based on the confusion matrix and the receiver operation characteristic curve.Finally,the explanable model is used to carry out an experiment on tourism volume forecasting.The experimental results show that the proposed model not only has high accuracy,but also is suitable for short-term and seasonal tourism volume forecasting.