Capital Flow Prediction Of Yu'e Bao Under Big Data Risk Management

With the enactment of the white paper “The new generation of artificial intelligence development plan” by the state council in 2017,the development of artificial intelligence is included in the national strategy.Financial technology,as an essential application under the theme of “artificial intelligence” and “big data”,has attracted much attention.The core challenge for financial firms is maximizing commercial profits while reducing investment risks.Considering the liquidity risk and surplus as the main factors in the company's investment strategy,to put forward a scheme about cash flow forecast,is important to meet the financial product of risk control demands.The traditional financial forecast mainly uses qualitative methods to guide theoretically,which is subjective and hard to generalize.The groovelike quantitative methods use simplex linear models to predict the result.Although it relatively improves the scientific basis of the qualitative analysis,yet the prediction accuracy is not so satisfying and the theoretical foundation was relatively weak.Based on the qualitative discussion of rich financial theory,quantitative methods are adopted to predict the purchase and redemption amount.The model uses 2.8 million records of the Yu'e Bao users.This paper builds a traditional linear model called YEB_ARIMA and a modern YEB_LSTM model under artificial intelligence and deep learning,in order to predict the Yu'e Bao's captical flow.Moreover,an ensemble leaning method,including linear combination and logistic regression learning,is applied to improve the original weak classifiers.The YEB_Hybrid model achieves 84.39% and 84.36% accuracy on the purchase and redemption dataset.Finally,according to the result of capital forecast,the proposal of fund reserve ratio is given.The model using big data analysis performs well under a variety of evaluation indexes,which exceeds other forecast methods on Yu 'e Bao capital flow.The main work and innovation points of this paper are explained as follows:1?In order to predict the fund flow of Yu'e Bao more accurately,the essay analyzes the dataset provided by Ant Financial Service Group(Alibaba)and combines the theory of time series analysis to do research.By obtaining the dataset features,more accurate models can be further built.2?Qualitative analysis is taken into concideration as well as quantitative analysis,and linear methods are combined with nonlinear ones.Through much theoretical background qualitative analysis,we combine multiple modeling methods to research on time series.For quantitative modeling methods,the linear model and nonlinear model are both adopted to observe the dataset from different aspects.The linear method uses YEB_ARIMA model.This model determines the model arguments by the ACF and PACF parameters,and then chooses the optimal model parameters by the tuning method Gridsearch.The nonlinear method uses the current popular YEB_LSTM long short term memory network to achieve good prediction results.3?Ensemble learning is introduced to obtain a higher level of prediction model.The linear model YEB_ARIMA and the nonlinear model YEB_LSTM are combined by linear combination method and stacking learning method to achieve YEB_Hybrid.The different models observe the data from various aspects,luckily,the final prediction effect is improved.4?Multiple evaluation methods are adopted to improve the reliability of the model.The models are checked by noise characteristic,Ljung-Box Q-test,and error calculation.The models are also compared with the existing models.The algorithm proposed is recognized as feasible and effective.This method can be used to predict the purchase and redemption amount.Then a reasonable proportion plan is given.This paper carries out the idea of ”technology is the wisdom and application is kindom”,which relies on artificial intelligence and big data to explore the application scenarios of new technologies in the field of financial management.We will implement smart finance into daily life and provide new impetus for future business development.5?On the basis of accurate predictions of the fund flow,a fund reserve ratio is further computed.The risk control suggestion is then raised.Applying Fintech to practice is deeply responsive to the call of the new generation of artificial intelligence development plan in the country,which will make great contribution.