关键词: 移动支付, 大数据决策, 风险控制, 机器学习, 特征选择

Abstract: With the rapid development of mobile payment industry, the security of mobile payment has become an issue of great concern to the society. The arrival of the era of big data makes the establishment of risk identification model of mobile payment by means of big data become a feasible method to guarantee the security of mobile payment. There are two major problems in the process of using the thirdparty platform of mobile payment for data modeling: First, the timeliness and stability of the model should be guaranteed. Let the machine learning model learn the latest mobile payment risks and avoid repeated training as long as possible. Secondly, the risk control system of the thirdparty payment platform will intervene in many dangerous transactions based on the risk judgment of transactions and generate unlabeled data. How to use these unlabeled data is also a problem to be solved. The experimental results showed that the smooth feature is selected from these data can improve the stability of the model by comparing the statistical features of the training set and the test set data, and labeling the unlabeled data as negative samples can also improve the model effect.Key wordsmobile payment; big data decision; risk control; machine learning; feature selection

Key words: mobile payment, big data decision, risk control, machine learning, feature selection