Research On Key Technology Of Bank Term Deposit Prediction Based On Machine Learning

Due to the maturity of the Internet industry,the competitive launch of Internet financial derivatives,and the change in people’s consumption attitudes,the deposit business of the banking industry has been severely impacted.The research on the problem of bank term deposit Prediction can help the bank to deal with the crisis of the deposit business.This paper mainly studies the key technologies in the prediction of bank term deposits,proposes a feature selection method based on the improved chicken swarm algorithm,and builds an ensemble model for bank term deposit prediction based on the Stacking method.The main work includes:1)Aiming at the feature selection problem,this paper proposes a feature selection method based on the Improved Chicken Swarm Optimization(ICSO)algorithm.The algorithm introduces the Levy flight strategy in the position update of the hen and adds the nonlinear strategy to reduce the inertia weight and the influence of the rooster behavior in the position update of the chick,so as to enhance the global search ability of the algorithm and avoid the algorithm from falling into the local optimal solution.In order to verify the effectiveness of the algorithm in this paper,it is compared with other algorithms on 18 UCI data sets,and the results show that the algorithm in this paper has obvious advantages.Likewise,the feature selection method improves the accuracy of the bank term deposit classification model.2)Unlike previous studies that only used a single model as a classification model,in order to predict bank term deposits,this paper constructs an ensemble learning model based on the Stacking method.This model is composed of 2 layers of models stacked,the first layer is composed of 4 basic learners: decision tree,multi-layer perceptron,random forest,and GBDT;the second layer is composed of only one meta-learner logistic regression.The first layer is trained using the full training set,and the second layer is trained using a 5-fold cross-validation method.3)This paper uses the Portuguese bank dataset Bank Marketing of the UCI platform to train the model.On the basis of an in-depth understanding of the bank term deposit prediction business,through exploratory analysis of the dataset,SMOTE algorithm to deal with category imbalance,and feature selection based on the ICSO algorithm,10 feature variables were selected to train the model.In order to verify the superiority of the prediction ability of the ensemble model proposed in this paper,it is not only compared with the single model of the four base learners but also with the models based on the Bagging method,the Boosting method,the hard voting method,and the soft voting method.The experimental results show that the ensemble model proposed in this paper is higher than several other comparable models in the three evaluation indicators of accuracy,F1 score,and AUC.From this,it can be concluded that the ensemble model based on the Stacking method proposed in this paper has an excellent performance in the prediction of bank term deposits.