Abstract:Currently, technological innovation has been increasingly valued by companies around the world, and patents as an important recording medium of innovation show a tendency for explosive growth. They are used to perform the classification of technical documents that are obtained with different predefined categories. However the presence of numerous patent documents obtained as a challenging task in classification and not able to convey effective knowledge. So this study deals with automatic annotation and classification techniques of patents has intrigued many scholars. First, this paper is based on the analysis of patent text features, including application number, publication number, application date, publication date, classification number, applicant, inventor, title, abstract, claims, full text, and the like. Second, we selected patent text fields, proposed to classify three tags namely technical field, technical effect, and technical means by the title, abstract, specification, background, and full-text field of the patent, and described its effectiveness. Then, we trained three different models Text Convolutional neural network (CNN), Bidirectional Long short term memory (Bi-LSTM), and ALBERT-Tiny, and chose Bi-LSTM and then multiple factors are employed to validate processing time and accuracy. Moreover, Alpine Skiing Optimization (ASO) is an initial search strategy that is determined for feature classification, focusing on selecting relevant features for analysis. Conclusively, the classification results of technical field, technical effect, and technical means based on the parameters trained using the Bi-LSTM model, and the prediction accuracy were 88.39%, 88.67%, and 88.53%, respectively. The existing methods can be compared with proposed with the performance metrics accuracy, precision, and recall which attained higher performance with 98.01%, 97%, and 98.56%. This model can satisfy actual needs, and its accuracy can be improved through further enrichment of features.

Application of SVM-KNN Combination Improvement Algorithm on Patent Text Classification

An automatic classification method for patents

A novel pattern recognition algorithm: Combining ART network with SVM to reconstruct a multi-class classifier

SVM-KNN Classifier--A New Method of Improving the Accuracy of SVM Classifier

Performance Analysis and Comparison of Neural Networks and Support Vector Machines Classifier

A research on text categorization based on the fusion of K-NN and SVM

MSVM-kNN: Combining SVM and K-Nn for Multi-class Text Classification

An Improved Support Vector Classification and Its Applicationn

Two-level hierarchical combination method for text classification

Enhancing patent text classification with Bi-LSTM technique and alpine skiing optimization for improved diagnostic accuracy

SVMDT Classifier and Its Application Research in Text Classification

A text classification method based on improved KNN algorithm

An Algorithm Based on KNN and Improved SVM for Licence Plate Recognition

SVM Classification:Its Contents and Challenges

Improved KNN Algorithm based on Probability and Adaptive K Value.

An Ameliorated Pattern Classification Method of SVM

Patent Automatic Classification Based on Symmetric Hierarchical Convolution Neural Network

Classification Algorithm Research Based on Support Vector Machine and Reverse K-nearest Neighbor

A CNN-based Feature Extraction Scheme for Patent Analysis

A Non-VSM kNN algorithm for text classification

Research on Feature Selection and Knn Classification Method in Chinese Text Classification