隨著國內不少優秀的棒球選手進入美國職棒聯盟，而且運動彩券也於民國97年正式發行，除了創造專業運動評論的需求，也吸引了國人對於運動彩券投注的興趣，因此我們希望透過分析歷史資料來預測棒球賽事。 本論文中，我們從球隊得分的觀點切入，建立棒球賽事的預測模型，期望透過預測兩支球隊的得分，來推論賽事結果，並可當作投注運彩的策略。 我們的預測模型結合了類神經網路與迴歸分析的理論，首先透過類神經網路去預測球隊的打擊表現，接著利用迴歸分析的技術，並參考Pete Palmer提出的Batting Runs公式，為每支球隊建立專屬的得分公式，然後將預測的打擊表現套用得分公式，計算球隊的預測得分。最後根據預測結果來模擬投注棒球運動彩券，並分析報酬率。 實作時，我們採用美國棒球聯盟近三年之資料來測試我們的方法，實驗結果顯示，我們預測的結果能獲得不錯的報酬率。 / There are many outstanding Taiwanese baseball players who joined the MLB in the USA, and the sport lottery has been issued in 2008. These have not only generated the demand for professional commentaries on various sports but also attracted numerous interests in sport lottery playing. Hence, we hope to predict the future baseball game results through analyzing the historical data. In this thesis, we started from the scoring mechanism and developed a model to predict baseball games. We used the predicted results, together with the sports lottery playing regulations, to find the lottery winning strategies. We used artificial neural networks and regression analysis in our model. Through the neural networks we can predict the batting behaviors of each team. Then, we used the regression analysis and the batting runs formula, proposed by Pete Palmer, to establish the scoring formula for each team. Combining the batting behaviors and the scoring formula, we can predict the scores of the future baseball games. Finally, we used the predicted scores to estimate the winning rate of sports lottery on these baseball games. We used the materials in the past three years of MLB in USA to verify our method. The experimental results show that we can obtain good winning rate.