摘要
樂透號碼預測，在實務上與學術上都是一項有趣且重要的議題。實務上，坊間存在許多各式各樣的樂透號碼預測方式，例如，連莊、拖牌、專車、立柱、時事牌等，但這些預測策略的效用如何，並無定論，也缺少實證資料的支持。學術上，首先關心的是樂透號碼「能否預測」這項議題，如果統計隨機假設在樂透的開獎號碼上成立，則樂透預測似乎成了緣木求魚，這類研究的成果多支持隨機假設，但結論還無法完全為人所接受。因此，也有部分研究實際利用一些簡單的分析方法或資訊技術，來進行樂透號碼的預測。本論文先擱置樂透號碼「能否預測」這項議題，重心放在驗證當代的資訊科技技術是否能夠改善樂透預測的效果。具體而言，本研究利用資料探勘中的頻繁樣式分析技術與貪婪演算法，來實作坊間一些常見的猜牌策略，主要包括拖牌、專車與立柱策略，並以臺灣彩劵公司的開獎紀錄，實際驗證資訊科技結合傳統猜牌策略的樂透預測效果，並以先前研究較少進行的獲利率評估方式，來衡量我們提的預測方法的效能。根據本論文的實驗結果，要讓系統預測的樂透號碼穩定地得到正獲利率是有一定難度的，但在某些特定條件下，還是有機會讓樂透購買結果賺錢。
並列摘要
Lottery numbers prediction is an interesting and important topic in practice and academic. In practice, lottery players are interested in methods that help to foretell the winning numbers. There are various kinds of prediction algorithms, such as analysis of repetition pattern, analysis of hot-cold trend, analysis of group, analysis of adjoining pairs, beed proposed. On the other hand, academic researches generally concern the topic of randomness of lottery winning numbers. Some researches try to prove that lottery winning numbers fit the randomess assumption, while others attempt to find out non-randomness patterns of lottery bets. In this study, we concentrate on using emerging information technology to implement well-known lottery prediction algorithms and then empirically evaluate their effectiveness. Specifically, we adopt the ideas of frequent pattern mining techniques (i.e., sequental pattern mining and association rule mining) to discover frequent subsequent and concurrent patterns of lottery numbers. Subsequently, two purchasing strategies are conducted. According to our empirical evaluation results using Taiwan lottery data, ranged form January 5, 2004 to Februry 8, 2013, our porposed prediction method has some chances to win the lottery jackpot in the cost of over 90% average lost rate.