A personalized academic literature recommendation method, including the following steps: S1 data collection and cleaning: collecting paper data with papers and authors as the core, the paper data including paper titles, paper abstracts, author names, publication years, publications and references, cleaning out data with obvious format errors and missing data; S2 model establishment, the process is as follows: S2.1 Construction of training set; S2.2 Feature calculation; S3 model training; S4 Academic literature recommendation, the process is as follows: S4 .1 To establish a set of candidate documents, it is required that the publication time of the cited paper selected in each step is earlier than the publication time of the paper; S4.2 Forecasting, the paper with the highest k′ in the probability value is taken as the final recommended reference. The invention can more accurately and efficiently generate a list of references meeting user needs.