随着金融市场的发展,金融产品的种类越来越多,消费者有了更多的选择。为了分析消费者对金融产品的选择并做出相关预测,一家金融数据咨询公司评估了客户对金融数据服务和数据库系统的需求和满意度。本文基于评估调查数据,分别采用随机森...随着金融市场的发展,金融产品的种类越来越多,消费者有了更多的选择。为了分析消费者对金融产品的选择并做出相关预测,一家金融数据咨询公司评估了客户对金融数据服务和数据库系统的需求和满意度。本文基于评估调查数据,分别采用随机森林分类算法(RF)、Adaboost分类算法、K近邻(KNN)分类算法建立预测模型;同时,针对每种模型开展特征重要性分析,探究不同因素对金融产品选择的影响程度。结果表明,K近邻(KNN)的分类模型的预测能力最佳;三种模型提供的可解释性基本符合实际规律,且特征重要性排序规律定性基本一致:存储指标2、使用时间指标和数据库大小对消费者选择金融产品影响显著,有无国家经济数据所占比例最低。With the development of the financial market, there are more and more types of financial products, and consumers have more choices. In order to analyze consumers’ choice of financial products and make relevant predictions, a financial data consulting company assesses customers’ demand and satisfaction with financial data services and database systems. Based on the survey data, Adaboost classification algorithm, K-nearest neighbor (KNN) classification algorithm and random forest classification algorithm (RF) were used to establish prediction models. At the same time, we carry out feature importance analysis for each model to explore the influence of different factors on the selection of financial products. The results show that K-nearest neighbor (KNN) classification model has the best prediction ability. The interpretability provided by the three models basically conforms to the actual law, and the characteristics of the importance ranking law are basically the same qualitatively: storage index 2, the use time index and the size of the database have a significant impact on consumers’ choice of financial products, and the proportion of whether there is national economic data is the lowest.展开更多
文摘随着金融市场的发展,金融产品的种类越来越多,消费者有了更多的选择。为了分析消费者对金融产品的选择并做出相关预测,一家金融数据咨询公司评估了客户对金融数据服务和数据库系统的需求和满意度。本文基于评估调查数据,分别采用随机森林分类算法(RF)、Adaboost分类算法、K近邻(KNN)分类算法建立预测模型;同时,针对每种模型开展特征重要性分析,探究不同因素对金融产品选择的影响程度。结果表明,K近邻(KNN)的分类模型的预测能力最佳;三种模型提供的可解释性基本符合实际规律,且特征重要性排序规律定性基本一致:存储指标2、使用时间指标和数据库大小对消费者选择金融产品影响显著,有无国家经济数据所占比例最低。With the development of the financial market, there are more and more types of financial products, and consumers have more choices. In order to analyze consumers’ choice of financial products and make relevant predictions, a financial data consulting company assesses customers’ demand and satisfaction with financial data services and database systems. Based on the survey data, Adaboost classification algorithm, K-nearest neighbor (KNN) classification algorithm and random forest classification algorithm (RF) were used to establish prediction models. At the same time, we carry out feature importance analysis for each model to explore the influence of different factors on the selection of financial products. The results show that K-nearest neighbor (KNN) classification model has the best prediction ability. The interpretability provided by the three models basically conforms to the actual law, and the characteristics of the importance ranking law are basically the same qualitatively: storage index 2, the use time index and the size of the database have a significant impact on consumers’ choice of financial products, and the proportion of whether there is national economic data is the lowest.
