Time-series Multi-view Clustering Method for Process Route Discovery

The determination and extraction of typical process routes are crucial to improving production efficiency and quality control in industrial production. Traditional process route extraction methods generally rely on similarity calculations and clustering analysis, exhibiting limited multi-view data adaptability and computational efficiency. To resolve these issues, a multi-view subspace clustering algorithm that incorporates temporal-smoothness constraints and adaptive weighting is proposed. This algorithm employs adaptive weighting to adjust data representation across different views and combines dynamic time warping to effectively align time series data, thereby achieving precise process route extraction. This method demonstrates excellent clustering performance on various real-world datasets across multiple parameters and effectively addresses the volatility and uncertainty of time series data. The experimental results reveal that the proposed algorithm effectively improves clustering accuracy and can significantly address complex multi-view data and time series features, thereby reflecting its innovativeness in multi-view and time series clustering analysis.