為因應選舉法規變更或時代變遷，往往必須重劃選區。傳統上，選區重劃都是以人工方式劃分。以人工方式劃分選區固然能考慮較多因素，包括最難數據化的人文因素，但人力成本高，也容易引起爭議。 本研究中，我們提出一個有系統的方式以自動劃分選區。主要的考慮因素為選區之人口數、選區形狀及二級行政區之完整性。我們的劃分方式主要分為三部份：產生起始選區、二級行政區分割修正及選區形狀調整。在產生起始選區步驟，我們根據位能場的觀念，劃分出人口數符合標準之起始選區，再經過行政區分割修正以維持二級行政區之完整性，最後採用基因演算法來調整選區形狀，以避免傑利蠑螈的狀況。 我們以台北市為例，來闡述我們的方法，實做的結果顯示我們的方法能有效的做正確的選區重劃。 / Electoral redistricting is normally required when election regulations changed. Traditionally, electoral redistricting is done manually. Though manual redistricting could consider humane or cultural factor, which may be very difficult to be included in the computation model, the cost of manual redistricting normally is high. In addition, manual redistricting may induce controversial issues. In this thesis, we propose a systematic way that could do the electoral redistricting automatically. Our major considerations are: (1) the population must be evenly partitioned, within an acceptable error; (2) the shape of the redistricted region is reasonably good; (3) the integrity of the second level district must be kept reasonably well. Our method consists of three major parts: initial district production, district’s integrity fixing, and district reshaping. The concept of potential is used in producing the initial districts. A heuristic is used in fixing the district’s integrity. And, finally, Genetic Algorithm is used in district reshaping. We use Taipei City as an example to illustrate our idea. Experimental results show that our method can do electoral redistricting effectively.