In semiconductor manufacturing, each wafer comprises multiple chips, and each chip is tested before the packaging process. Wafer test data on electrical characteristics of chips are collected during the wafer test process. However, missing values often occur due to various manufacturing environments. In this study, a new missing value imputation method based on Generative Adversarial Imputation Nets (GAIN) is proposed. The proposed method takes into account the two characteristics of wafer test data, namely, spatial similarity among chips and test item correlation. Spatial similarity refers to the property of having similar test item values between chips in adjacent or symmetrical positions. Test item correlation refers to the positive correlation between test items with similar physical properties. Spatial similarity and test item correlation are reflected by the addition of locational information of chips and modification of the loss function in GAIN, respectively. The performance of the proposed method is validated with a real wafer test dataset by a comparison with those of existing methods in various circumstances.