Patents contain detailed information about the developed technology. In addition, patents grant exclusive rights to the developed technology. For this reason, many companies use patents for technology protection. Patent litigation occurs when the operating activities of one company infringe on the scope of the patent rights of another company. When patent litigation occurs, a lot of time and money are consumed. Therefore, it is necessary to prevent patent litigation in advance. In this paper, an appropriate feature extraction method is sought when constructing a model for classifying patent litigation. Principal component analysis and Autoencoder are used to perform the proposed research. The experimental data are those registered with the USPTO as patents related to artificial intelligence. Feature extraction is performed on the quantitative indicators of the collected patents. In addition, performance is measured with various classification algorithms. As a result of the experiment, the classification performance of the method combining Autoencoder and K-Nearest neighbor was good.