Application of Machine Learning Algorithms in Project Economics Review

Abstract

In order to solve the problem that the traditional query based on secondary retrieval is too rigid so as to automatically filter out valuable target documents, and repeated queries consume a lot of time, an intelligent interactive information retrieval process and processing flow for fund project document queries is proposed. Based on the feedback information of users evaluating project documents, ID3 algorithm, CLCC algorithm and SVM classification function are used to learn the potential intention and target of users' query respectively, and the learned rule knowledge or classification function is applied to support the project document query. The experimental computations and analysis are conducted for the query of project documents in a fund review management as an example. The results show that the number of project documents read and evaluated by the user in each interactive query loop is no more than 5% of the total number of documents returned from the previous query or 20 items, and together with the project documents already read and evaluated, they constitute the set of machine learning samples. The maximum number of interactive query cycle is set to 5. Among the three machine learning methods, ID3 also shows good prediction performance when post-processing algorithm is used. ID3 generates a decision tree with large width and small height. CLCC algorithm is better than ID3 algorithm, mainly because the rule post-processing of CLCC is more flexible, and the generated concept rules contain more merge rules and each merge rule is shorter. The SVM method has the best prediction performance, mainly because the project document keyword vectors are all continuous real values. It is concluded that the fund project intelligent interactive information retrieval process and processing flow accurately describes the potential query intention and target of user evaluation project documents, and establishes a user query project document classification learning knowledge base system, thus realizing the knowledge-based project document query support.