Patent lifetime prediction using LightGBM with a customized loss

• DOI: 10.7717/peerj-cs.2044
• Publication Date: 2024-05-10T07:21:11Z
• Authors: Jieming Liu; Peizhao Li; Xiaowei Liu
• Keywords: Patent lifetime; Algorithms and Analysis of Algorithms; Electronic computers. Computer science; Machine learning; 0202 electrical engineering, electronic engineering, information engineering; QA75.5-76.95; 02 engineering and technology; Prediction; LightGBM; Data science

Patent lifespan is commonly used as a quantitative measure in patent assessments. holders maintain exclusive rights by paying significant maintenance fees, suggesting strong correlation between patent’s and its business potential or economic value. Therefore, accurately forecasting the duration of great significance. This study introduces highly effective method that combines LightGBM, sophisticated machine learning algorithm, with customized loss function derived from Focal Loss. The purpose this approach to predict probability remaining valid until maximum expiration date. research differs previous studies have examined various stages phases patents. Instead, it assesses commercial viability individual patents considering their lifespan. evaluation process utilizes dataset consisting 200,000 experimental results show improvement performance model combining Loss LightGBM. By incorporating into ability give priority difficult instances during training enhanced, resulting an overall performance. targeted enhances model’s distinguish different samples recover challenges giving samples. As result, improves accuracy making predictions apply those new data.