งานวิจัย
\"การพยากรณ์หลักทรัพย์โดยใช้เทคนิคการค้นหากฎความสัมพันธ์\"
โดย :
ดร.อารยา อริยา
อ.ชิดชนก วงศ์เครือ
อ.จิระประภา คำราช
อ.ประเสริฐ ยังปากน้ำ
| ชื่อเรื่อง : | การพยากรณ์หลักทรัพย์โดยใช้เทคนิคการค้นหากฎความสัมพันธ์ |
| ผู้วิจัย : | ดร.อารยา อริยา อ.ชิดชนก วงศ์เครือ อ.จิระประภา คำราช อ.ประเสริฐ ยังปากน้ำ |
| วันที่เผยแพร่ : | 30-11-2017 |
| บทคัดย่อ : | |
| Abstract : | Stock Forecasting using Association Rule Discovery Araya Ariya1 Chitchanok Wongkhrua2 Prasert Youngpaknam3 Jiraprapa Kamratch4 1araya_aa@hotmail.com, 2chitcha_24@hotmail.com, 3satool_2504@hotmail.co.th, 4chi_2514@hotmail.com Lampang Rajabhat University, Lampang, Thailand Abstract: Stock market is an important part of the economy in each country. Many researchers focus on the accuracy prediction of stock trend. In this paper, the association rule mining is applied to forecast stock price. Data used is 10 stocks from Stock Exchange of Thailand (SET) website. All data are 6 months collected. There are 6 experiments in this research; there are 1 to 6 months of collected data. Each experiment is test with 4 minimum supports: 10%, 20%, 30% and 40%. The experiment results show that the investor should analyze at least 4 stocks for monitoring trend of stock price. Keywords: Association rules Discovery, Stock Forecasting, Data mining 1. INTRODUCTION An association rule mining, one of the important tasks in data mining, is a well-known research topic that many researchers purposed a large number of algorithms for solving association rule discovering problems. This problem was first presented by Agrawal et al [1] which analyzes the behavior of customer purchasing to help business make a decision. The result showed co-occurrence buying items called frequent patterns, which can generate interesting rules. From that research, a problem statement of an association rule mining is defined as follows. Let I = {I1,I2,…,Im} be a set of literal items. DB is a database which contains transactions. Each transaction T is a set of items where T I. Given X is an item and X I. Each transaction contain X if and only if X T. Let X and Y are an item where X I, Y I and X Y0. Each set of items , itemsets, is called a frequent itemset if and only if its support is greater than or equal to support threshold s%. It calculates from a number of transactions in DB that contain X Y. An association rule can be showed in XY form. Each frequent itemset can be made the association rule if and only if it is satisfied by confidence threshold c% which calculates from a number of transactions in DB that contain X and also contain Y. Both s% and c% are specified by user. After an association rule mining was revealed, it motivated many researchers to extend this research area in a lot of issues such as fraud detection, bioinformatics, economics, educational, etc. In this paper, the association rule mining is applied to forecast stock price. The remaining of this paper, the literatures of an association rule mining is reviewed in section 2. Data preparation is described in section 3. An applying association rule discovery to forecast shared stock and its experiment are presented in section 4 and 5 respectively. The paper conclusion and future work are briefed in section 6. 2. RELATED WORK Mining association rules was first purposed by Agrawal et al [1] which finds a correlation between itemsets in a transaction database. The finding rules process has 2 steps: finding frequent itemsets (sometimes they are called large itemsets) and generating rules. Subsequent years, Apriori [2], the most popular algorithm, was purposed to discover frequent itemsets. Generally, databases are always appended. Scanning a whole database for k-passes, Apriori is unsuitable for mining association rules in dynamic database. From this situation, existing valid rules may be invalid. Cheung et al [3] purposed fast update algorithm, FUP, to solve a rules maintenance problem by using the previous knowledge to find frequent itemsets in updated database. The concept of FUP is re-using frequent itemsets of previous mining to update with frequent itemsets of an incremental database. Although FUP can decrease a number of candidate itemsets for scanning original database, it still needs to scan an original database k times when new frequent itemsets are found. This can degrade the performance of FUP algorithm. In our observation, the advantage of FUP are re-using frequent itemsets of a previous mining to prune itemsets which cannot be a frequent itemset in updated database and reducing candidate itemsets to scan in an original database. However, the disadvantage of FUP is the algorithm needs to scan an original database equal to a size of k frequent itemsets, e.g., if maximum size of k frequent itemsets is k=5, FUP needs to scan an original database for 5 times. From that problem, re-scanning an original database, this paper proposes an incremental association rule mining algorithm to improve the performance of FUP algorithm by reducing a number of a scanning times of an original database. 4. REFERENCES [1] Agrawal R, Imielinski T, Swami A (1993), A mining association rules between sets of items in large database, In Proceeding of the ACM SIGMOD Int’l Conf. on Management of Data (ACM SIGMOD’93), Washington, USA, May 1993, pp.207-216. [2] Agrawal R, Srikant R (1994), Fast algorithm for mining association rules, In Proc. 20th Int. Conf. Very Large DataBases (VLDB’94), Santiago, Chile, September 12-15, 1994, pp.487-499. [3] Cheung D.W., Han J, Ng V.T., Wong C.Y., Maintenance of Discovered Association rules in Large Databases: An incremental updating technique, In 12th IEEE International Conference on Data Engineering, pp 106-114, 1996. [4] Tsai Paulry S.M., Lee Chih-Chong, Chen Abree L.P. (2005), An Efficient Approach for Incremental Association Rule Mining, Proceedings of the third Pacific-Asia Conference on Methodologies for Knowledge Discovery and Data Mining, Lecture Notes In Computer Science, Vol. 1574 archieve, 1999. |
| รายละเอียด : | |