以倒傳遞網路設計籃球運動彩券推薦模式

Using Back Propagation Network to Design Basketball Sports Lottery Recommendation Model

陳贊仁(Tsan-Jen Chen)

指導教授 ： 黃有評 ； 謝尚琳

大同大學/電機資訊學院/資訊工程學系所/碩士(2009年)

摘要

近年來運動彩券在市場上逐漸的成熟，但至今都沒有針對運動彩券所開發的推薦系統。一般在現實生活中，運動彩券分析師會針對特定賽事透過自我的專業能力，給予彩券迷賽事推薦的資訊，但準確度普遍都不高。因此，本論文將以籃球項目為例之運動彩券，設計一套推薦系統。
本系統利用球隊及球員歷史數據，根據交戰紀錄、主客場及球員近況因素等，探勘出有價值的資訊，以便預測出比賽結果，並針對使用者在網頁上所點選的賽事進行推薦。預測機制首先是以資料探勘策略來取得有效賽事及有效賽況數據。本論文利用數據複雜度計算出該場比賽最具有影響力的球員作為有效的賽況數據。此外，以倒傳遞類神經網路針對前置處理所取得的有效賽事細部資料進行訓練。
在預測賽事方面，本論文是透過主客兩隊前一場交戰紀錄來取得球隊數據，並計算出最具有影響力的球員。本論文利用灰預測技術找出球員近況表現趨勢、透過近五場各項數據得知球員近況表現以及該球員過去歷史細部數據來得知球員優劣勢能力。將取得的各項預測值再以倒傳遞類神經網路預測出其賽事結果。此外，本論文針對不同類型的玩法產生出兩種不同的倒傳遞類神經網路。
本論文實驗結果顯示讓分、不讓分及大小分三種玩法其準確度分別為64.04%、71.91%及65.17%。此結果證明，本論文利用資料探勘策略、倒傳遞類神經網路及灰預測技術可有效掌握賽事，並可針對籃球所有賽事於運動彩券不同玩法上進行推薦。
關鍵詞：運動彩券、資料探勘、倒傳遞類神經網路、推薦機制、灰預測。

關鍵字

灰預測 ； 推薦機制 ； 倒傳遞類神經網路 ； 資料探勘 ； 運動彩券

並列摘要

In recent years, sports lottery gradually matured in the market. But so far there is no feasible recommendation model for the sports lottery. In real life, sports lottery professional analysts based on their expertise to give users the recommendation information. But, the recommendation information is generally not accurate enough. Therefore, this thesis uses basketball as an example to design a recommendation model for sports lottery.
The proposed system considers each player’s and his team’s historical records in matching performace, home and guest palying scores to mine valuable information and to predict the matching results that are in turn used for the users to reference when clicking the recommendation games on the website. The prediction module is based on data mining strategy to obtain representative games and effective game data. This thesis uses combined data to find the most influential players for matching games. Back-propagation neural network is applied to preprocess effective data during the training phase.
We obtained the team data through previous host and guest matched game records, which are then used to find the most influential players. The trend of performance predicted by gray model, the performance of recent five games, and the past records are combined to judge the strength and weakness of players. The overall evaluation data are used by back-propagation network to predict the game results. Besides, this thesis proposes two back-propagation networks for different types of playing methods.
The experimental results show that the game accuracies for spread, non-spread and totals are 64.04%, 71.91% and 65.17%, respectively. The results verify that by using data minimg, back-propagation network and gray prediction model can track the game tempo, and the proposed system can be used for recommendation to NBA games.
Keywords: Sports lottery, data mining, back-propagation network, recommendation mechanism, grey prediction.

並列關鍵字

data mining ； back-propagation network ； recommendation mechanism ； Sports lottery ； grey prediction

參考文獻

[4] S.Y. Kang, A Study on Awareness and Purchase Intention of Sports Lottery in Taiwan, Graduate Institute of Physical Education, Fu Jen Catholic University, June 2008.

[8] W.N. Pi and L.C. Liou, “Electric power demand forecasting in taiwan via grey prediction,” Journal of Science and Engineering Technology, vol. 3, no. 2, pp.11-18, 2007.

[12] 洪國禎、吳國榮，改良式GM(1,1)灰預測模型於台電電量需求預測之研究，國防大學運籌管理學系，工程科技教育學刊，vol. 5, no. 3, pp.446-458, 2008年9月。

[5] K. Park, M. Bell, I. Kaparias and K. Bogenberger, “Learning user preferences of route choice behaviour for adaptive route guidance,” IET Intelligent Transport Systems, vol. 1, no. 2, pp.159-166, June 2007.

[6] S. Park, S. Kang and Y.K. Kim, “A channel recommendation system in mobile environment,” IEEE Trans. on Consumer Electronics, vol. 52, no. 1, pp.33-39, Feb. 2006.