There are three types of error in Taiwan's election prediction: errors from survey process, transplanting of measure scale, and no response vot-er. This paper try to reduce the error of election prediction via "Joined Id-iosyncrasies Adjusted Model" (JIA Model). JIA Model was operated by two stages. First, I compute a basic model, which reflects some general factors in every county. Secondly, I design extended model to adjust the output of basic model. In JIA Model, I try to use logistic regression to compute the candidate's ballots, and present the final prediction in proba-bility. In this paper, I try to explain in example of 1997 Taipei Mayoral Election. I consider the factors of issue orientation, strategic voting, so-cial context and political map. JIA Model made the error less than sam-pling errors.