Scholarlink

<p>Fuzzy time series (FTS) models are widely used in prediction tasks due to their ability to handle uncertain and nonlinear problems effectively. However, type-1 fuzzy sets have limitations in dealing with data noise and linguistic ambiguity, particularly in multifactor situations. This study proposes a multifactor stock index forecasting model based on type-2 fuzzy sets and a variable size interval partitioning technique. Firstly, information granules are constructed with enhanced effectiveness and interpretability. This is achieved using the fuzzy <i>C</i>-means (FCM) clustering algorithm and the principle of justifiable granularity to partition the universe of discourse. The FCM algorithm in this model attains optimal clustering results by employing an interval type-2 fuzzy approach to adjust the fuzzy parameter <i>m</i> in FCM. Secondly, to more accurately capture the non-determinacy in the time series, three observations of the stock index are modeled as type-2 fuzzy sets, which results in a group of refined first-order fuzzy relationships. To evaluate the performance of this model, experiments were conducted using six benchmark datasets from the Taiwan Stock Exchange Index (TAIEX). The corresponding root mean square error (RMSE) was calculated as the evaluation criterion. Compared with nine representative time series forecasting approaches, the experimental results demonstrate that our model not only guarantees effectiveness and robustness but also achieves enhanced prediction performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A novel multifactor type-2 fuzzy time series model based on improved fuzzy C-means algorithm and justifiable granularity for stock index forecasting

  • 期刊论文

  • 发表日期：

    2025年4月1日

查看全文
  • Zengtai Gong,
  • Jindong Feng

摘要

Fuzzy time series (FTS) models are widely used in prediction tasks due to their ability to handle uncertain and nonlinear problems effectively. However, type-1 fuzzy sets have limitations in dealing with data noise and linguistic ambiguity, particularly in multifactor situations. This study proposes a multifactor stock index forecasting model based on type-2 fuzzy sets and a variable size interval partitioning technique. Firstly, information granules are constructed with enhanced effectiveness and interpretability. This is achieved using the fuzzy C-means (FCM) clustering algorithm and the principle of justifiable granularity to partition the universe of discourse. The FCM algorithm in this model attains optimal clustering results by employing an interval type-2 fuzzy approach to adjust the fuzzy parameter m in FCM. Secondly, to more accurately capture the non-determinacy in the time series, three observations of the stock index are modeled as type-2 fuzzy sets, which results in a group of refined first-order fuzzy relationships. To evaluate the performance of this model, experiments were conducted using six benchmark datasets from the Taiwan Stock Exchange Index (TAIEX). The corresponding root mean square error (RMSE) was calculated as the evaluation criterion. Compared with nine representative time series forecasting approaches, the experimental results demonstrate that our model not only guarantees effectiveness and robustness but also achieves enhanced prediction performance.