Study Of Intelligent Investment Recommendation With Missing Data

With the flourishing development of science and technology,the traditional financial industry is facing profound changes,and the intelligence of investment advisory services is one of the manifestations of the application of Fintech technology to financial product recommendations and portfolio management.However,although the recommendation quality of intelligent investment models based on machine learning is getting higher and higher,the following problems still exist in general.First,while using the latest technology to improve the accuracy of results,the importance of model interpretability for financial decision making has been neglected,and little derivation and analysis of investor preferences has been done.Second,only static forms of data can be handled effectively,but investors’ transaction data are often dynamic in nature.Third,there is a lack of in-depth research on how to deal with the problem of missing data in a timely and efficient manner.To address the current problems of insufficient financial significance and inability to effectively handle dynamic data and missing data in intelligent investment recommendation models,this paper develops an intelligent investment recommendation method in the case of missing data based on online learning and group decision making technique.First,by combining investor preference derivation with the parameter of logistic regression models,we provide financial explanations for the models from the perspective of investor preferences.Second,by demonstrating that the loss function of the logistic regression model are convex,logistic regression can be extended to online optimization.By giving the upper bound of the objective function of the online logistic regression model,we show that the online learning method can find a set of optimal solutions of the model with high efficiency,which is beneficial to solve the problem of dynamic generation of financial data.In addition,by using the introduction of matrices in the objective function to describe the relationship between data features and by reflecting the magnitude of influence among investors through networks,data missing processing methods applicable to a variety of data missing contexts are developed,and the effectiveness of both data missing processing methods is demonstrated in both the mathematical basis and the actual dataset.Therefore,the innovations of this paper can be roughly divided into the following aspects: First,the new model avoids the dilemma that results are difficult to interpret due to the complex mechanism of the underlying machine learning models,which is conducive to gaining users’ trust and has certain guiding significance for investors’ behavioral decisions.Second,by extending the algorithm from the mathematical level,we develop an online learning form of logistic regression model and prove its effectiveness.Other machine learning models have complex loss functions that are difficult to meet the mathematical and theoretical requirements of online learning methods.The efficiency improvement from the underlying optimization remains a non-negligible advantage of the new model.Third,the temporal validity of the data missing processing methods developed from different perspectives is not only mathematically proven,but also the enhancement effect on the model results has passed the test of the actual data set.Fourth,based on the above,the application mode of science and technology in intelligent investment recommendation is further extended by providing a recommendation system with the new model in this paper.