The existing K-line patterns are acquired by artificial observation, i. e., artificial similarity search. Hence,there are a series of problems that the forecast performance of these patterns is actually modest, and some researchers even deny their possibility. Based on these problems, the restudy of K-line series prediction based on similarity search was presented using the methods of computer technology and data mining. Firstly, the similarity measure model was defined to solve the similarity match problem of K-line series, including the shape similarity model and the position similarity model.Secondly, based on the similarity measure model, the K-line sliding search algorithm was proposed to resolve the problem of K-line series＇ similarity search. Finally, based on the similarity search results, two stock prediction methods were presented,which are common series based similarity search and pattern series based similarity search. In the experiment, the forecast accuracies of the methods of common series and K-line patterns can reach to 72. 5% and 77. 8%, respectively. The experimental results show that, the K-line patterns do have predictive ability, and its predictive performance is better than common series. The proposed two prediction methods could be well applied in stock prediction and investment.
journal of Computer Applications