自適應專家策略在雙向拍賣之研究

Adaptive Specialist Strategy in Double Auctions

余榮裕(Jeng-Yu Yu)

指導教授 ： 陳燈能

國立屏東科技大學/管理學院/資訊管理系所/碩士(2009年)

https://doi.org/10.6346/NPUST.2009.00204

摘要

本研究探討連續雙向拍賣(Continuous Double Auctions, CDA)的市場
經營策略之效率問題,並將其導入台灣果菜批發市場之中,探討雙向拍
賣市場機制對於農產品供需價格所產生的影響。
本研究首先分析藉交易代理人競賽(Trading Agent Competition, TAC)
之雙向拍賣市場設計競賽(TAC Market Design Competition)所產生的市場
交易模擬資料,以雙向拍賣市場經營者(專家)面向,探討經營雙向拍賣
市場的各項市場交易政策對於交易代理人的影響,並分析拍賣市場經營
績效指標,包括市佔率、獲利與成交率彼此之間的關聯,提出改善雙向
拍賣市場經營政策之方法,並建構一套具有自主適應競爭態勢能力的市
場經營專家代理人,用以自動化調節市場政策,提升市場經營績效。
本研究並採用 2007 及 2008 年台灣食用甜玉米拍賣交易價格數據,
模擬現實環境參數並進行雙向拍賣,比較採用雙向拍賣市場機制與現行
批發市場所採用的拍賣方式,對於玉米供需價格所造成的差異,結果發
現本研究所提出之市場交易機制具有穩定供需價格的功用,證明了雙向
拍賣市場對於在農產品交易上所具有的優勢。

關鍵字

雙向拍賣 ； 市場設計 ； 智慧代理人 ； 奈許平衡 ； 決策樹 ； K 最近鄰居法 ； 果菜批發市場

並列摘要

In this research, we study Continuous Double Auctions (CDA)
mechanism and investigate the CDA market design strategies to gain better
market performance. In advance, we introduce CDA to Taiwan Fruit and
Vegetable Wholesale Markets to analyze if CDA would affect on the trading
price.
Trading Agent Competition (TAC) is a worldwide famous competition
for applying intelligent agent technology to simulate business operation, and
TAC Market Design Competition is one of TAC competitions that focus on
the CDA market. In this research, we collect and analyze the transaction data
in the TAC to identify the impacts of market design strategies on the market
performance, such as market share, market profit and transaction success rate.
Based on the analysis results, we develop our adaptive CDA market design
strategy to fit the dynamic transaction environment and enlarge market
performance.
Finally, we apply our adaptive CDA market design strategy to analyze
Taiwan eating corn transaction data in 2007 and 2008. The results show that
our market design has better stability on the transaction price. It would help
the famers to get fair trade in the market.

並列關鍵字

Double Auction ； Market Design ； Intelligent Agent ； Nash Equilibrium ； Decision Tree ； K-Nearest Neighbour ； Fruit and Vegetable Wholesale Market

參考文獻

Report ULCS-09-005, Department of Computer Science, University of

3. Cover, T. (1968). Estimation by the nearest neighbor rule. Information

Theory, IEEE Transactions on, 14(1), 50-55.

Reinforcement Learning in Experimental Games with Unique, Mixed

Strategy Equilibria. American Economic Review, American Economic