Classification method of multi-modal intelligence data

Abstract

The invention discloses a multi-modal intelligence data classification method, and relates to the field of artificial intelligence data processing. Comprising the following steps: converting multi-source data file contents in various formats into text data, and constructing an original text data set; performing data merging, rule filtering and comparison reduction operation on the text data obtained by conversion to realize data denoising; supplementing the cleaned data missing, and extracting keywords and abstracts of the text data; and adopting a classification method based on a TFIDF weight improvement algorithm to obtain a classification result of the text data. Scientific and technological information data are further subdivided into key parts such as titles, keywords and abstracts, the TFIDF weight is adjusted by considering the word frequency of the words in class titles and class keywords, the special positions of the words are considered, the classification capacity of the words on classes is enhanced, and the classification accuracy of the classification model on scientific and technological information texts can be improved.

Images